diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..ef5f64faff8041ba0b806a7d500e804e6c3a283f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: transformers
+model_name: Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1
+tags:
+- generated_from_trainer
+- trl
+- sft
+licence: license
+---
+
+# Model Card for Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1
+
+This model is a fine-tuned version of [Qwen/Qwen3.5-4B-Base](https://huggingface.co/Qwen/Qwen3.5-4B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/katriin-kukk/Cross_lingual_morphological_generalization/runs/mq7in2f1) 
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 0.29.0
+- Transformers: 5.5.4
+- Pytorch: 2.10.0
+- Datasets: 4.6.1
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+    
+```bibtex
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
+}
+```
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..28c9af374285086f1c66ae341c34b93e41de0cba
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/README.md
@@ -0,0 +1,58 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: transformers
+model_name: Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2
+tags:
+- generated_from_trainer
+- trl
+- sft
+licence: license
+---
+
+# Model Card for Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2
+
+This model is a fine-tuned version of [Qwen/Qwen3.5-4B-Base](https://huggingface.co/Qwen/Qwen3.5-4B-Base).
+It has been trained using [TRL](https://github.com/huggingface/trl).
+
+## Quick start
+
+```python
+from transformers import pipeline
+
+question = "If you had a time machine, but could only go to the past or the future once and never return, which would you choose and why?"
+generator = pipeline("text-generation", model="None", device="cuda")
+output = generator([{"role": "user", "content": question}], max_new_tokens=128, return_full_text=False)[0]
+print(output["generated_text"])
+```
+
+## Training procedure
+
+[<img src="https://raw.githubusercontent.com/wandb/assets/main/wandb-github-badge-28.svg" alt="Visualize in Weights & Biases" width="150" height="24"/>](https://wandb.ai/katriin-kukk/Cross_lingual_morphological_generalization/runs/gtp4i6ta) 
+
+
+
+This model was trained with SFT.
+
+### Framework versions
+
+- TRL: 0.29.0
+- Transformers: 5.5.4
+- Pytorch: 2.10.0
+- Datasets: 4.6.1
+- Tokenizers: 0.22.2
+
+## Citations
+
+
+
+Cite TRL as:
+    
+```bibtex
+@software{vonwerra2020trl,
+  title   = {{TRL: Transformers Reinforcement Learning}},
+  author  = {von Werra, Leandro and Belkada, Younes and Tunstall, Lewis and Beeching, Edward and Thrush, Tristan and Lambert, Nathan and Huang, Shengyi and Rasul, Kashif and Gallouédec, Quentin},
+  license = {Apache-2.0},
+  url     = {https://github.com/huggingface/trl},
+  year    = {2020}
+}
+```
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e2c0cb87592557b2e0b777b18355d538da4a1e9
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-100/trainer_state.json
@@ -0,0 +1,139 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.24140012070006034,
+  "eval_steps": 20,
+  "global_step": 100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9897324005793792.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d52a66f922983baf4b7708bc8c1c62a0fa065562
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1000/trainer_state.json
@@ -0,0 +1,1084 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.4103802051901027,
+  "eval_steps": 20,
+  "global_step": 1000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.8038608775637e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..74975bdc6174ea63fb70de63db1e142d8ccd0e76
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1020/trainer_state.json
@@ -0,0 +1,1105 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.4586602293301145,
+  "eval_steps": 20,
+  "global_step": 1020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0007168784646349e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..294ebe3a07f37c7f2253819b28a70ea82df8bbe4
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1040/trainer_state.json
@@ -0,0 +1,1126 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.506940253470127,
+  "eval_steps": 20,
+  "global_step": 1040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0198081900655616e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..407be8578f8f9715455c10f0b50df2715d7de8b8
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1060/trainer_state.json
@@ -0,0 +1,1147 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.5552202776101387,
+  "eval_steps": 20,
+  "global_step": 1060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0385749388429107e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c3f7c0059398e17d28cbe3ad17c1cd561472cb6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1080/trainer_state.json
@@ -0,0 +1,1168 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.603500301750151,
+  "eval_steps": 20,
+  "global_step": 1080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0583776570652467e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bee2e41c45a36d30435ba9eab57e5700ac4512aa
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1100/trainer_state.json
@@ -0,0 +1,1189 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.651780325890163,
+  "eval_steps": 20,
+  "global_step": 1100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0779277426032845e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b8775179929192efb85de85d21760ce0884e54b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1120/trainer_state.json
@@ -0,0 +1,1210 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.700060350030175,
+  "eval_steps": 20,
+  "global_step": 1120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.0975532670678835e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..53cac2f79c5b82f613cb0f75ddbd22ca96571a85
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1140/trainer_state.json
@@ -0,0 +1,1231 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.748340374170187,
+  "eval_steps": 20,
+  "global_step": 1140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1159542247710925e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c87d762141a7ca7e179c91a3a3f2c2a475632d32
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1160/trainer_state.json
@@ -0,0 +1,1252 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.796620398310199,
+  "eval_steps": 20,
+  "global_step": 1160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1362964190380237e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f67e8446b62e27bae6624608038e8c9842d5bbb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1180/trainer_state.json
@@ -0,0 +1,1273 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.8449004224502112,
+  "eval_steps": 20,
+  "global_step": 1180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1553096601219277e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a9c3c2a531dd004c4b0ec82483eefda4b865d50
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-120/trainer_state.json
@@ -0,0 +1,160 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.28968014484007243,
+  "eval_steps": 20,
+  "global_step": 120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2000140223848448e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..81ecc3a28e9e1407929bb80072b219927f537c56
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1200/trainer_state.json
@@ -0,0 +1,1294 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.8931804465902236,
+  "eval_steps": 20,
+  "global_step": 1200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.1750869396364698e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a50eb773303a5c057ee27e02074e41c1c7ae1ad0
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1220/trainer_state.json
@@ -0,0 +1,1315 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.9414604707302354,
+  "eval_steps": 20,
+  "global_step": 1220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.19498878109952e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e80a97e70bcb9c97ed1adfc8263c4a8e3372a57d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1240/trainer_state.json
@@ -0,0 +1,1336 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.9897404948702473,
+  "eval_steps": 20,
+  "global_step": 1240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2137441263182029e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..60f547b242c87199ebf25b18d08591c5d551e709
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1260/trainer_state.json
@@ -0,0 +1,1357 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.036210018105009,
+  "eval_steps": 20,
+  "global_step": 1260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2326518844844442e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..332cecf04279532d3c8bad77cdcbe8d1a65bc11f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1280/trainer_state.json
@@ -0,0 +1,1378 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.084490042245021,
+  "eval_steps": 20,
+  "global_step": 1280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2543756637436314e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5ad5e937d3c4b84b22af8f8bbe25ce68aa31ac3
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1300/trainer_state.json
@@ -0,0 +1,1399 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.1327700663850333,
+  "eval_steps": 20,
+  "global_step": 1300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2752494392837734e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac3478328063f859c71862d70ca61c441f8b1a44
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1320/trainer_state.json
@@ -0,0 +1,1420 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.181050090525045,
+  "eval_steps": 20,
+  "global_step": 1320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.2964977778662605e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba125efad50d1dd7176776b1d1e3424dea108738
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1340/trainer_state.json
@@ -0,0 +1,1441 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.2293301146650575,
+  "eval_steps": 20,
+  "global_step": 1340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3153899657889382e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5194ff4aa85ee5959c570cb49e4ceb70a5db8edc
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1360/trainer_state.json
@@ -0,0 +1,1462 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.2776101388050694,
+  "eval_steps": 20,
+  "global_step": 1360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3352751405123994e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6489d837cef993c2efc54f9b41fc570cce69fcf4
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1380/trainer_state.json
@@ -0,0 +1,1483 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.3258901629450817,
+  "eval_steps": 20,
+  "global_step": 1380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.353556799448023e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc2c17decc6786466fd63b700fd8ca8f03690aae
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-140/trainer_state.json
@@ -0,0 +1,181 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.33796016898008446,
+  "eval_steps": 20,
+  "global_step": 140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4166816370434048e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..088c5fb9bd16bcc536119229d62d2f6314a24f0a
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1400/trainer_state.json
@@ -0,0 +1,1504 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.3741701870850935,
+  "eval_steps": 20,
+  "global_step": 1400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3721270561375846e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..71e74a25bf49138f382a82fc16e0c2eb5f530274
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1420/trainer_state.json
@@ -0,0 +1,1525 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.4224502112251054,
+  "eval_steps": 20,
+  "global_step": 1420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.3915490709398323e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..305009f07e968518d6de4b373df9873868119f74
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1440/trainer_state.json
@@ -0,0 +1,1546 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.4707302353651177,
+  "eval_steps": 20,
+  "global_step": 1440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4106009086838374e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..af128f47c793515e14b4c70d689a28e87ad10dbd
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1460/trainer_state.json
@@ -0,0 +1,1567 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.51901025950513,
+  "eval_steps": 20,
+  "global_step": 1460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.430731698517033e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..13b41f75062153b0f1c01a850003bea2a6ba7022
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1480/trainer_state.json
@@ -0,0 +1,1588 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.567290283645142,
+  "eval_steps": 20,
+  "global_step": 1480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4509668747718656e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b98d411894115a8fa8bc216dc86ddabfb7025101
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1500/trainer_state.json
@@ -0,0 +1,1609 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.6155703077851538,
+  "eval_steps": 20,
+  "global_step": 1500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.470379240409088e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..66c6a384b4e38adb7abb59653387a0ccba85d082
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1520/trainer_state.json
@@ -0,0 +1,1630 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.663850331925166,
+  "eval_steps": 20,
+  "global_step": 1520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.4890284448124928e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7e064d67f1fbbf5793c017e69c9b8f0a1034adb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1540/trainer_state.json
@@ -0,0 +1,1651 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.712130356065178,
+  "eval_steps": 20,
+  "global_step": 1540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5089864268720538e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9ba6e1c4a38cf05f02040cfea74c9c2d1bb784e
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1560/trainer_state.json
@@ -0,0 +1,1672 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.7604103802051903,
+  "eval_steps": 20,
+  "global_step": 1560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.528082124457083e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..744c0b8d7ab996f1fb39b7d087fb4241a2537a4b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1580/trainer_state.json
@@ -0,0 +1,1693 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.808690404345202,
+  "eval_steps": 20,
+  "global_step": 1580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5468058905902285e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..945b18aac792ef942e90e133d4739d17aa451936
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-160/trainer_state.json
@@ -0,0 +1,202 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.38624019312009655,
+  "eval_steps": 20,
+  "global_step": 160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6047701792907264e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..76cc434ab65d0ff04d4d0774a02924e1c105df04
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1600/trainer_state.json
@@ -0,0 +1,1714 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.856970428485214,
+  "eval_steps": 20,
+  "global_step": 1600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5667805393893786e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f962b4ccb042f123aaffc5aa64cb48b8210afd9b
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1620/trainer_state.json
@@ -0,0 +1,1735 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.9052504526252263,
+  "eval_steps": 20,
+  "global_step": 1620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.5864385201363354e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b7066690474ff355bac5901d964cc22c5d0f15f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1640/trainer_state.json
@@ -0,0 +1,1756 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.9535304767652386,
+  "eval_steps": 20,
+  "global_step": 1640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.606152641479803e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..387cccee220087408d27b2ec3ce922eff3d66f50
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1660/trainer_state.json
@@ -0,0 +1,1777 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.0,
+  "eval_steps": 20,
+  "global_step": 1660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6246204854405734e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab02a9e1ce73480530d9d9e58512c9ed34bcef60
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1680/trainer_state.json
@@ -0,0 +1,1798 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.048280024140012,
+  "eval_steps": 20,
+  "global_step": 1680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.644855661695406e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c0f7d8241c409a662b07b4bfbe4a956051cf784
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1700/trainer_state.json
@@ -0,0 +1,1819 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.096560048280024,
+  "eval_steps": 20,
+  "global_step": 1700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6634952169337856e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d7b22ed986ce139fecd3992aed232e418099a96
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1720/trainer_state.json
@@ -0,0 +1,1840 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.1448400724200365,
+  "eval_steps": 20,
+  "global_step": 1720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.6840961842627584e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5626e5417a0fc82f910c2c0472e3ef432be532f
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1740/trainer_state.json
@@ -0,0 +1,1861 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.193120096560048,
+  "eval_steps": 20,
+  "global_step": 1740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7031883730605056e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b138c0e56e1c1d97219a5d9fb051d1f21117a1ec
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1760/trainer_state.json
@@ -0,0 +1,1882 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.24140012070006,
+  "eval_steps": 20,
+  "global_step": 1760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.722359509572096e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b96299fefa9b92a089c1706ac63a07ceaf9dbd42
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1780/trainer_state.json
@@ -0,0 +1,1903 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.289680144840072,
+  "eval_steps": 20,
+  "global_step": 1780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7419517005575168e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..160ec58aa6af5df5cb61b415e493c10d3777adfa
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-180/trainer_state.json
@@ -0,0 +1,223 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.43452021726010864,
+  "eval_steps": 20,
+  "global_step": 180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.800621913399296e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..374b1f7f68c07f0d737a6c168d446bc832940831
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1800/trainer_state.json
@@ -0,0 +1,1924 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.337960168980085,
+  "eval_steps": 20,
+  "global_step": 1800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7626254752225894e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3101dab1100823853b186c862869418deebe1e1
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1820/trainer_state.json
@@ -0,0 +1,1945 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.386240193120097,
+  "eval_steps": 20,
+  "global_step": 1820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    },
+    {
+      "entropy": 0.4086425334215164,
+      "epoch": 4.386240193120097,
+      "grad_norm": 0.8936498761177063,
+      "learning_rate": 0.0002596489848011036,
+      "loss": 0.3710860013961792,
+      "mean_token_accuracy": 0.8773891530930996,
+      "num_tokens": 4216243.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386240193120097,
+      "eval_entropy": 0.49018637685293565,
+      "eval_loss": 0.6762362718582153,
+      "eval_mean_token_accuracy": 0.8242661363623115,
+      "eval_num_tokens": 4216243.0,
+      "eval_runtime": 88.3684,
+      "eval_samples_per_second": 16.069,
+      "eval_steps_per_second": 2.014,
+      "step": 1820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.7815615229862912e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b11139a3ec8dbc8169cee2d837f4c12ced4f7a5
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1840/trainer_state.json
@@ -0,0 +1,1966 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.434520217260109,
+  "eval_steps": 20,
+  "global_step": 1840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    },
+    {
+      "entropy": 0.4086425334215164,
+      "epoch": 4.386240193120097,
+      "grad_norm": 0.8936498761177063,
+      "learning_rate": 0.0002596489848011036,
+      "loss": 0.3710860013961792,
+      "mean_token_accuracy": 0.8773891530930996,
+      "num_tokens": 4216243.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386240193120097,
+      "eval_entropy": 0.49018637685293565,
+      "eval_loss": 0.6762362718582153,
+      "eval_mean_token_accuracy": 0.8242661363623115,
+      "eval_num_tokens": 4216243.0,
+      "eval_runtime": 88.3684,
+      "eval_samples_per_second": 16.069,
+      "eval_steps_per_second": 2.014,
+      "step": 1820
+    },
+    {
+      "entropy": 0.41956189945340155,
+      "epoch": 4.434520217260109,
+      "grad_norm": 0.6450079083442688,
+      "learning_rate": 0.00025671146992703106,
+      "loss": 0.3715682029724121,
+      "mean_token_accuracy": 0.878043319284916,
+      "num_tokens": 4262463.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434520217260109,
+      "eval_entropy": 0.4821142643020394,
+      "eval_loss": 0.6749269962310791,
+      "eval_mean_token_accuracy": 0.8252766048640348,
+      "eval_num_tokens": 4262463.0,
+      "eval_runtime": 87.6946,
+      "eval_samples_per_second": 16.193,
+      "eval_steps_per_second": 2.03,
+      "step": 1840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.8004984479468134e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..63c36d787fe7b62ca08e522b886a1db21d8574e5
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1860/trainer_state.json
@@ -0,0 +1,1987 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.4828002414001205,
+  "eval_steps": 20,
+  "global_step": 1860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    },
+    {
+      "entropy": 0.4086425334215164,
+      "epoch": 4.386240193120097,
+      "grad_norm": 0.8936498761177063,
+      "learning_rate": 0.0002596489848011036,
+      "loss": 0.3710860013961792,
+      "mean_token_accuracy": 0.8773891530930996,
+      "num_tokens": 4216243.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386240193120097,
+      "eval_entropy": 0.49018637685293565,
+      "eval_loss": 0.6762362718582153,
+      "eval_mean_token_accuracy": 0.8242661363623115,
+      "eval_num_tokens": 4216243.0,
+      "eval_runtime": 88.3684,
+      "eval_samples_per_second": 16.069,
+      "eval_steps_per_second": 2.014,
+      "step": 1820
+    },
+    {
+      "entropy": 0.41956189945340155,
+      "epoch": 4.434520217260109,
+      "grad_norm": 0.6450079083442688,
+      "learning_rate": 0.00025671146992703106,
+      "loss": 0.3715682029724121,
+      "mean_token_accuracy": 0.878043319284916,
+      "num_tokens": 4262463.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434520217260109,
+      "eval_entropy": 0.4821142643020394,
+      "eval_loss": 0.6749269962310791,
+      "eval_mean_token_accuracy": 0.8252766048640348,
+      "eval_num_tokens": 4262463.0,
+      "eval_runtime": 87.6946,
+      "eval_samples_per_second": 16.193,
+      "eval_steps_per_second": 2.03,
+      "step": 1840
+    },
+    {
+      "entropy": 0.40638086050748823,
+      "epoch": 4.4828002414001205,
+      "grad_norm": 0.8167530298233032,
+      "learning_rate": 0.0002537545510911465,
+      "loss": 0.3652146100997925,
+      "mean_token_accuracy": 0.8761600315570831,
+      "num_tokens": 4311451.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4828002414001205,
+      "eval_entropy": 0.5017492558514134,
+      "eval_loss": 0.6696512699127197,
+      "eval_mean_token_accuracy": 0.8252955695216575,
+      "eval_num_tokens": 4311451.0,
+      "eval_runtime": 88.02,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.8212880125921894e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b3f73ffd8e04d83a8575f22feb039e93600738a6
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1880/trainer_state.json
@@ -0,0 +1,2008 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.531080265540133,
+  "eval_steps": 20,
+  "global_step": 1880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    },
+    {
+      "entropy": 0.4086425334215164,
+      "epoch": 4.386240193120097,
+      "grad_norm": 0.8936498761177063,
+      "learning_rate": 0.0002596489848011036,
+      "loss": 0.3710860013961792,
+      "mean_token_accuracy": 0.8773891530930996,
+      "num_tokens": 4216243.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386240193120097,
+      "eval_entropy": 0.49018637685293565,
+      "eval_loss": 0.6762362718582153,
+      "eval_mean_token_accuracy": 0.8242661363623115,
+      "eval_num_tokens": 4216243.0,
+      "eval_runtime": 88.3684,
+      "eval_samples_per_second": 16.069,
+      "eval_steps_per_second": 2.014,
+      "step": 1820
+    },
+    {
+      "entropy": 0.41956189945340155,
+      "epoch": 4.434520217260109,
+      "grad_norm": 0.6450079083442688,
+      "learning_rate": 0.00025671146992703106,
+      "loss": 0.3715682029724121,
+      "mean_token_accuracy": 0.878043319284916,
+      "num_tokens": 4262463.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434520217260109,
+      "eval_entropy": 0.4821142643020394,
+      "eval_loss": 0.6749269962310791,
+      "eval_mean_token_accuracy": 0.8252766048640348,
+      "eval_num_tokens": 4262463.0,
+      "eval_runtime": 87.6946,
+      "eval_samples_per_second": 16.193,
+      "eval_steps_per_second": 2.03,
+      "step": 1840
+    },
+    {
+      "entropy": 0.40638086050748823,
+      "epoch": 4.4828002414001205,
+      "grad_norm": 0.8167530298233032,
+      "learning_rate": 0.0002537545510911465,
+      "loss": 0.3652146100997925,
+      "mean_token_accuracy": 0.8761600315570831,
+      "num_tokens": 4311451.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4828002414001205,
+      "eval_entropy": 0.5017492558514134,
+      "eval_loss": 0.6696512699127197,
+      "eval_mean_token_accuracy": 0.8252955695216575,
+      "eval_num_tokens": 4311451.0,
+      "eval_runtime": 88.02,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1860
+    },
+    {
+      "entropy": 0.41664722077548505,
+      "epoch": 4.531080265540133,
+      "grad_norm": 0.757738471031189,
+      "learning_rate": 0.0002507790650665036,
+      "loss": 0.368887734413147,
+      "mean_token_accuracy": 0.8786893397569656,
+      "num_tokens": 4357724.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.531080265540133,
+      "eval_entropy": 0.4798511176296834,
+      "eval_loss": 0.6738249659538269,
+      "eval_mean_token_accuracy": 0.8250340461061242,
+      "eval_num_tokens": 4357724.0,
+      "eval_runtime": 88.2713,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.8415258204374835e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ad930305095fca12c7b63763edb7b4d5bf760c5
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1900/trainer_state.json
@@ -0,0 +1,2029 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.579360289680145,
+  "eval_steps": 20,
+  "global_step": 1900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    },
+    {
+      "entropy": 0.4086425334215164,
+      "epoch": 4.386240193120097,
+      "grad_norm": 0.8936498761177063,
+      "learning_rate": 0.0002596489848011036,
+      "loss": 0.3710860013961792,
+      "mean_token_accuracy": 0.8773891530930996,
+      "num_tokens": 4216243.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386240193120097,
+      "eval_entropy": 0.49018637685293565,
+      "eval_loss": 0.6762362718582153,
+      "eval_mean_token_accuracy": 0.8242661363623115,
+      "eval_num_tokens": 4216243.0,
+      "eval_runtime": 88.3684,
+      "eval_samples_per_second": 16.069,
+      "eval_steps_per_second": 2.014,
+      "step": 1820
+    },
+    {
+      "entropy": 0.41956189945340155,
+      "epoch": 4.434520217260109,
+      "grad_norm": 0.6450079083442688,
+      "learning_rate": 0.00025671146992703106,
+      "loss": 0.3715682029724121,
+      "mean_token_accuracy": 0.878043319284916,
+      "num_tokens": 4262463.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434520217260109,
+      "eval_entropy": 0.4821142643020394,
+      "eval_loss": 0.6749269962310791,
+      "eval_mean_token_accuracy": 0.8252766048640348,
+      "eval_num_tokens": 4262463.0,
+      "eval_runtime": 87.6946,
+      "eval_samples_per_second": 16.193,
+      "eval_steps_per_second": 2.03,
+      "step": 1840
+    },
+    {
+      "entropy": 0.40638086050748823,
+      "epoch": 4.4828002414001205,
+      "grad_norm": 0.8167530298233032,
+      "learning_rate": 0.0002537545510911465,
+      "loss": 0.3652146100997925,
+      "mean_token_accuracy": 0.8761600315570831,
+      "num_tokens": 4311451.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4828002414001205,
+      "eval_entropy": 0.5017492558514134,
+      "eval_loss": 0.6696512699127197,
+      "eval_mean_token_accuracy": 0.8252955695216575,
+      "eval_num_tokens": 4311451.0,
+      "eval_runtime": 88.02,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1860
+    },
+    {
+      "entropy": 0.41664722077548505,
+      "epoch": 4.531080265540133,
+      "grad_norm": 0.757738471031189,
+      "learning_rate": 0.0002507790650665036,
+      "loss": 0.368887734413147,
+      "mean_token_accuracy": 0.8786893397569656,
+      "num_tokens": 4357724.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.531080265540133,
+      "eval_entropy": 0.4798511176296834,
+      "eval_loss": 0.6738249659538269,
+      "eval_mean_token_accuracy": 0.8250340461061242,
+      "eval_num_tokens": 4357724.0,
+      "eval_runtime": 88.2713,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1880
+    },
+    {
+      "entropy": 0.42732664197683334,
+      "epoch": 4.579360289680145,
+      "grad_norm": 0.7890474200248718,
+      "learning_rate": 0.00024778585388045044,
+      "loss": 0.38672595024108886,
+      "mean_token_accuracy": 0.8742976360023021,
+      "num_tokens": 4401934.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579360289680145,
+      "eval_entropy": 0.49321879295820603,
+      "eval_loss": 0.6752269268035889,
+      "eval_mean_token_accuracy": 0.824115453811174,
+      "eval_num_tokens": 4401934.0,
+      "eval_runtime": 87.3378,
+      "eval_samples_per_second": 16.259,
+      "eval_steps_per_second": 2.038,
+      "step": 1900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.8598951990551552e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0a9aa299943d8a6f9fb1e2115d0b402fdf15270
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1920/trainer_state.json
@@ -0,0 +1,2050 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.627640313820157,
+  "eval_steps": 20,
+  "global_step": 1920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    },
+    {
+      "entropy": 0.4086425334215164,
+      "epoch": 4.386240193120097,
+      "grad_norm": 0.8936498761177063,
+      "learning_rate": 0.0002596489848011036,
+      "loss": 0.3710860013961792,
+      "mean_token_accuracy": 0.8773891530930996,
+      "num_tokens": 4216243.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386240193120097,
+      "eval_entropy": 0.49018637685293565,
+      "eval_loss": 0.6762362718582153,
+      "eval_mean_token_accuracy": 0.8242661363623115,
+      "eval_num_tokens": 4216243.0,
+      "eval_runtime": 88.3684,
+      "eval_samples_per_second": 16.069,
+      "eval_steps_per_second": 2.014,
+      "step": 1820
+    },
+    {
+      "entropy": 0.41956189945340155,
+      "epoch": 4.434520217260109,
+      "grad_norm": 0.6450079083442688,
+      "learning_rate": 0.00025671146992703106,
+      "loss": 0.3715682029724121,
+      "mean_token_accuracy": 0.878043319284916,
+      "num_tokens": 4262463.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434520217260109,
+      "eval_entropy": 0.4821142643020394,
+      "eval_loss": 0.6749269962310791,
+      "eval_mean_token_accuracy": 0.8252766048640348,
+      "eval_num_tokens": 4262463.0,
+      "eval_runtime": 87.6946,
+      "eval_samples_per_second": 16.193,
+      "eval_steps_per_second": 2.03,
+      "step": 1840
+    },
+    {
+      "entropy": 0.40638086050748823,
+      "epoch": 4.4828002414001205,
+      "grad_norm": 0.8167530298233032,
+      "learning_rate": 0.0002537545510911465,
+      "loss": 0.3652146100997925,
+      "mean_token_accuracy": 0.8761600315570831,
+      "num_tokens": 4311451.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4828002414001205,
+      "eval_entropy": 0.5017492558514134,
+      "eval_loss": 0.6696512699127197,
+      "eval_mean_token_accuracy": 0.8252955695216575,
+      "eval_num_tokens": 4311451.0,
+      "eval_runtime": 88.02,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1860
+    },
+    {
+      "entropy": 0.41664722077548505,
+      "epoch": 4.531080265540133,
+      "grad_norm": 0.757738471031189,
+      "learning_rate": 0.0002507790650665036,
+      "loss": 0.368887734413147,
+      "mean_token_accuracy": 0.8786893397569656,
+      "num_tokens": 4357724.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.531080265540133,
+      "eval_entropy": 0.4798511176296834,
+      "eval_loss": 0.6738249659538269,
+      "eval_mean_token_accuracy": 0.8250340461061242,
+      "eval_num_tokens": 4357724.0,
+      "eval_runtime": 88.2713,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1880
+    },
+    {
+      "entropy": 0.42732664197683334,
+      "epoch": 4.579360289680145,
+      "grad_norm": 0.7890474200248718,
+      "learning_rate": 0.00024778585388045044,
+      "loss": 0.38672595024108886,
+      "mean_token_accuracy": 0.8742976360023021,
+      "num_tokens": 4401934.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579360289680145,
+      "eval_entropy": 0.49321879295820603,
+      "eval_loss": 0.6752269268035889,
+      "eval_mean_token_accuracy": 0.824115453811174,
+      "eval_num_tokens": 4401934.0,
+      "eval_runtime": 87.3378,
+      "eval_samples_per_second": 16.259,
+      "eval_steps_per_second": 2.038,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41300065964460375,
+      "epoch": 4.627640313820157,
+      "grad_norm": 0.866765022277832,
+      "learning_rate": 0.00024477576457634665,
+      "loss": 0.364247989654541,
+      "mean_token_accuracy": 0.8774793080985546,
+      "num_tokens": 4449745.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627640313820157,
+      "eval_entropy": 0.49179744352115673,
+      "eval_loss": 0.6731056571006775,
+      "eval_mean_token_accuracy": 0.8245729244826885,
+      "eval_num_tokens": 4449745.0,
+      "eval_runtime": 88.2458,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 1920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.8793058102987366e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..77b67615d29f9b5335eff7fbe0865f1549f13ba0
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1940/trainer_state.json
@@ -0,0 +1,2071 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.675920337960169,
+  "eval_steps": 20,
+  "global_step": 1940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    },
+    {
+      "entropy": 0.4086425334215164,
+      "epoch": 4.386240193120097,
+      "grad_norm": 0.8936498761177063,
+      "learning_rate": 0.0002596489848011036,
+      "loss": 0.3710860013961792,
+      "mean_token_accuracy": 0.8773891530930996,
+      "num_tokens": 4216243.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386240193120097,
+      "eval_entropy": 0.49018637685293565,
+      "eval_loss": 0.6762362718582153,
+      "eval_mean_token_accuracy": 0.8242661363623115,
+      "eval_num_tokens": 4216243.0,
+      "eval_runtime": 88.3684,
+      "eval_samples_per_second": 16.069,
+      "eval_steps_per_second": 2.014,
+      "step": 1820
+    },
+    {
+      "entropy": 0.41956189945340155,
+      "epoch": 4.434520217260109,
+      "grad_norm": 0.6450079083442688,
+      "learning_rate": 0.00025671146992703106,
+      "loss": 0.3715682029724121,
+      "mean_token_accuracy": 0.878043319284916,
+      "num_tokens": 4262463.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434520217260109,
+      "eval_entropy": 0.4821142643020394,
+      "eval_loss": 0.6749269962310791,
+      "eval_mean_token_accuracy": 0.8252766048640348,
+      "eval_num_tokens": 4262463.0,
+      "eval_runtime": 87.6946,
+      "eval_samples_per_second": 16.193,
+      "eval_steps_per_second": 2.03,
+      "step": 1840
+    },
+    {
+      "entropy": 0.40638086050748823,
+      "epoch": 4.4828002414001205,
+      "grad_norm": 0.8167530298233032,
+      "learning_rate": 0.0002537545510911465,
+      "loss": 0.3652146100997925,
+      "mean_token_accuracy": 0.8761600315570831,
+      "num_tokens": 4311451.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4828002414001205,
+      "eval_entropy": 0.5017492558514134,
+      "eval_loss": 0.6696512699127197,
+      "eval_mean_token_accuracy": 0.8252955695216575,
+      "eval_num_tokens": 4311451.0,
+      "eval_runtime": 88.02,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1860
+    },
+    {
+      "entropy": 0.41664722077548505,
+      "epoch": 4.531080265540133,
+      "grad_norm": 0.757738471031189,
+      "learning_rate": 0.0002507790650665036,
+      "loss": 0.368887734413147,
+      "mean_token_accuracy": 0.8786893397569656,
+      "num_tokens": 4357724.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.531080265540133,
+      "eval_entropy": 0.4798511176296834,
+      "eval_loss": 0.6738249659538269,
+      "eval_mean_token_accuracy": 0.8250340461061242,
+      "eval_num_tokens": 4357724.0,
+      "eval_runtime": 88.2713,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1880
+    },
+    {
+      "entropy": 0.42732664197683334,
+      "epoch": 4.579360289680145,
+      "grad_norm": 0.7890474200248718,
+      "learning_rate": 0.00024778585388045044,
+      "loss": 0.38672595024108886,
+      "mean_token_accuracy": 0.8742976360023021,
+      "num_tokens": 4401934.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579360289680145,
+      "eval_entropy": 0.49321879295820603,
+      "eval_loss": 0.6752269268035889,
+      "eval_mean_token_accuracy": 0.824115453811174,
+      "eval_num_tokens": 4401934.0,
+      "eval_runtime": 87.3378,
+      "eval_samples_per_second": 16.259,
+      "eval_steps_per_second": 2.038,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41300065964460375,
+      "epoch": 4.627640313820157,
+      "grad_norm": 0.866765022277832,
+      "learning_rate": 0.00024477576457634665,
+      "loss": 0.364247989654541,
+      "mean_token_accuracy": 0.8774793080985546,
+      "num_tokens": 4449745.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627640313820157,
+      "eval_entropy": 0.49179744352115673,
+      "eval_loss": 0.6731056571006775,
+      "eval_mean_token_accuracy": 0.8245729244826885,
+      "eval_num_tokens": 4449745.0,
+      "eval_runtime": 88.2458,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 1920
+    },
+    {
+      "entropy": 0.42646181732416155,
+      "epoch": 4.675920337960169,
+      "grad_norm": 0.7921772003173828,
+      "learning_rate": 0.00024174964897385928,
+      "loss": 0.3822937488555908,
+      "mean_token_accuracy": 0.8755100265145301,
+      "num_tokens": 4495074.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675920337960169,
+      "eval_entropy": 0.5194089940089858,
+      "eval_loss": 0.6604889631271362,
+      "eval_mean_token_accuracy": 0.8260471847619903,
+      "eval_num_tokens": 4495074.0,
+      "eval_runtime": 88.0139,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.898451508102533e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/README.md b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/adapter_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dd029c541eac295dd0ee9608fbb57ff26a95c6c
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 128,
+  "lora_bias": false,
+  "lora_dropout": 0.03828026524568501,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 64,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "gate_proj",
+    "v_proj",
+    "up_proj",
+    "down_proj",
+    "o_proj",
+    "q_proj",
+    "k_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/chat_template.jinja b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/tokenizer_config.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/trainer_state.json b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb81f5dc271f19ec233a805cdbda973c0357736d
--- /dev/null
+++ b/overgeneralisation_original_Estonian/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test2/checkpoint-1960/trainer_state.json
@@ -0,0 +1,2092 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.724200362100181,
+  "eval_steps": 20,
+  "global_step": 1960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 2.172999547421932,
+      "epoch": 0.04828002414001207,
+      "grad_norm": 2.5071208477020264,
+      "learning_rate": 1.7227585487708594e-05,
+      "loss": 1.9347333908081055,
+      "mean_token_accuracy": 0.6086816020309925,
+      "num_tokens": 47778.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.04828002414001207,
+      "eval_entropy": 1.6909754791956269,
+      "eval_loss": 1.4728385210037231,
+      "eval_mean_token_accuracy": 0.6773593797442619,
+      "eval_num_tokens": 47778.0,
+      "eval_runtime": 90.1265,
+      "eval_samples_per_second": 15.756,
+      "eval_steps_per_second": 1.975,
+      "step": 20
+    },
+    {
+      "entropy": 1.2014626950025558,
+      "epoch": 0.09656004828002414,
+      "grad_norm": 1.5074543952941895,
+      "learning_rate": 3.5361886001086065e-05,
+      "loss": 1.0867146492004394,
+      "mean_token_accuracy": 0.7285927519202232,
+      "num_tokens": 95981.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.09656004828002414,
+      "eval_entropy": 1.0304168352250302,
+      "eval_loss": 0.9211422204971313,
+      "eval_mean_token_accuracy": 0.7613122848312507,
+      "eval_num_tokens": 95981.0,
+      "eval_runtime": 88.6338,
+      "eval_samples_per_second": 16.021,
+      "eval_steps_per_second": 2.008,
+      "step": 40
+    },
+    {
+      "entropy": 0.9454198732972146,
+      "epoch": 0.14484007242003621,
+      "grad_norm": 1.7578108310699463,
+      "learning_rate": 5.349618651446354e-05,
+      "loss": 0.8447297096252442,
+      "mean_token_accuracy": 0.7723424732685089,
+      "num_tokens": 142784.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.14484007242003621,
+      "eval_entropy": 0.8647834474451086,
+      "eval_loss": 0.8240737318992615,
+      "eval_mean_token_accuracy": 0.7821227014064789,
+      "eval_num_tokens": 142784.0,
+      "eval_runtime": 88.0135,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 60
+    },
+    {
+      "entropy": 0.8963550612330436,
+      "epoch": 0.19312009656004828,
+      "grad_norm": 1.1464931964874268,
+      "learning_rate": 7.1630487027841e-05,
+      "loss": 0.7979836463928223,
+      "mean_token_accuracy": 0.7822489373385906,
+      "num_tokens": 187338.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.19312009656004828,
+      "eval_entropy": 0.8663296133614657,
+      "eval_loss": 0.7752988338470459,
+      "eval_mean_token_accuracy": 0.7885957012015782,
+      "eval_num_tokens": 187338.0,
+      "eval_runtime": 88.284,
+      "eval_samples_per_second": 16.084,
+      "eval_steps_per_second": 2.016,
+      "step": 80
+    },
+    {
+      "entropy": 0.8265103206038475,
+      "epoch": 0.24140012070006034,
+      "grad_norm": 1.0106662511825562,
+      "learning_rate": 8.976478754121848e-05,
+      "loss": 0.7297060012817382,
+      "mean_token_accuracy": 0.7955138415098191,
+      "num_tokens": 236038.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24140012070006034,
+      "eval_entropy": 0.8122169536151244,
+      "eval_loss": 0.754906177520752,
+      "eval_mean_token_accuracy": 0.7917155255092664,
+      "eval_num_tokens": 236038.0,
+      "eval_runtime": 87.5754,
+      "eval_samples_per_second": 16.215,
+      "eval_steps_per_second": 2.033,
+      "step": 100
+    },
+    {
+      "entropy": 0.8098392844200134,
+      "epoch": 0.28968014484007243,
+      "grad_norm": 1.2238041162490845,
+      "learning_rate": 0.00010789908805459595,
+      "loss": 0.7201489448547364,
+      "mean_token_accuracy": 0.7975016921758652,
+      "num_tokens": 282918.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.28968014484007243,
+      "eval_entropy": 0.8418726462326692,
+      "eval_loss": 0.7393178939819336,
+      "eval_mean_token_accuracy": 0.7936771585700217,
+      "eval_num_tokens": 282918.0,
+      "eval_runtime": 87.1597,
+      "eval_samples_per_second": 16.292,
+      "eval_steps_per_second": 2.042,
+      "step": 120
+    },
+    {
+      "entropy": 0.8080591082572937,
+      "epoch": 0.33796016898008446,
+      "grad_norm": 1.011591911315918,
+      "learning_rate": 0.00012603338856797343,
+      "loss": 0.7147697925567627,
+      "mean_token_accuracy": 0.7947175532579422,
+      "num_tokens": 331305.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.33796016898008446,
+      "eval_entropy": 0.7894941431083037,
+      "eval_loss": 0.720219075679779,
+      "eval_mean_token_accuracy": 0.801899236239744,
+      "eval_num_tokens": 331305.0,
+      "eval_runtime": 87.1493,
+      "eval_samples_per_second": 16.294,
+      "eval_steps_per_second": 2.042,
+      "step": 140
+    },
+    {
+      "entropy": 0.7957205109298229,
+      "epoch": 0.38624019312009655,
+      "grad_norm": 0.8481830954551697,
+      "learning_rate": 0.00014416768908135088,
+      "loss": 0.7050027847290039,
+      "mean_token_accuracy": 0.7995368830859662,
+      "num_tokens": 375876.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.38624019312009655,
+      "eval_entropy": 0.7746140591883928,
+      "eval_loss": 0.7175475358963013,
+      "eval_mean_token_accuracy": 0.7994848955213354,
+      "eval_num_tokens": 375876.0,
+      "eval_runtime": 86.655,
+      "eval_samples_per_second": 16.387,
+      "eval_steps_per_second": 2.054,
+      "step": 160
+    },
+    {
+      "entropy": 0.779565304517746,
+      "epoch": 0.43452021726010864,
+      "grad_norm": 1.032624363899231,
+      "learning_rate": 0.00016230198959472835,
+      "loss": 0.6978332042694092,
+      "mean_token_accuracy": 0.8015618294477462,
+      "num_tokens": 423913.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.43452021726010864,
+      "eval_entropy": 0.7455583871080634,
+      "eval_loss": 0.7118850350379944,
+      "eval_mean_token_accuracy": 0.7999682195400923,
+      "eval_num_tokens": 423913.0,
+      "eval_runtime": 86.7842,
+      "eval_samples_per_second": 16.362,
+      "eval_steps_per_second": 2.051,
+      "step": 180
+    },
+    {
+      "entropy": 0.7810172818601131,
+      "epoch": 0.4828002414001207,
+      "grad_norm": 0.9276663064956665,
+      "learning_rate": 0.00018043629010810582,
+      "loss": 0.6937861442565918,
+      "mean_token_accuracy": 0.802893053740263,
+      "num_tokens": 472395.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.4828002414001207,
+      "eval_entropy": 0.7650193290764027,
+      "eval_loss": 0.7020567059516907,
+      "eval_mean_token_accuracy": 0.8045302153973097,
+      "eval_num_tokens": 472395.0,
+      "eval_runtime": 88.326,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 200
+    },
+    {
+      "entropy": 0.7862060204148292,
+      "epoch": 0.5310802655401328,
+      "grad_norm": 0.8195295929908752,
+      "learning_rate": 0.0001985705906214833,
+      "loss": 0.6878782272338867,
+      "mean_token_accuracy": 0.8027168907225132,
+      "num_tokens": 517049.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.5310802655401328,
+      "eval_entropy": 0.7656565583154057,
+      "eval_loss": 0.6973932981491089,
+      "eval_mean_token_accuracy": 0.8047967278555538,
+      "eval_num_tokens": 517049.0,
+      "eval_runtime": 87.7698,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 220
+    },
+    {
+      "entropy": 0.7611534461379051,
+      "epoch": 0.5793602896801449,
+      "grad_norm": 0.7499712109565735,
+      "learning_rate": 0.00021670489113486077,
+      "loss": 0.681937837600708,
+      "mean_token_accuracy": 0.8041978515684605,
+      "num_tokens": 562836.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.5793602896801449,
+      "eval_entropy": 0.7599442380197933,
+      "eval_loss": 0.7002046704292297,
+      "eval_mean_token_accuracy": 0.8051861517884759,
+      "eval_num_tokens": 562836.0,
+      "eval_runtime": 87.9239,
+      "eval_samples_per_second": 16.15,
+      "eval_steps_per_second": 2.024,
+      "step": 240
+    },
+    {
+      "entropy": 0.7800783462822437,
+      "epoch": 0.627640313820157,
+      "grad_norm": 0.8645269274711609,
+      "learning_rate": 0.00023483919164823825,
+      "loss": 0.6954500675201416,
+      "mean_token_accuracy": 0.8031484372913837,
+      "num_tokens": 606612.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.627640313820157,
+      "eval_entropy": 0.7498679090751691,
+      "eval_loss": 0.7046149373054504,
+      "eval_mean_token_accuracy": 0.802388215667746,
+      "eval_num_tokens": 606612.0,
+      "eval_runtime": 88.2468,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 260
+    },
+    {
+      "entropy": 0.7586533665657044,
+      "epoch": 0.6759203379601689,
+      "grad_norm": 0.741596519947052,
+      "learning_rate": 0.0002529734921616157,
+      "loss": 0.690476369857788,
+      "mean_token_accuracy": 0.8046923100948333,
+      "num_tokens": 655920.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6759203379601689,
+      "eval_entropy": 0.7714274066887544,
+      "eval_loss": 0.6992160677909851,
+      "eval_mean_token_accuracy": 0.8042646482419432,
+      "eval_num_tokens": 655920.0,
+      "eval_runtime": 88.1754,
+      "eval_samples_per_second": 16.104,
+      "eval_steps_per_second": 2.019,
+      "step": 280
+    },
+    {
+      "entropy": 0.7585848286747933,
+      "epoch": 0.724200362100181,
+      "grad_norm": 1.0778124332427979,
+      "learning_rate": 0.00027110779267499317,
+      "loss": 0.6790409564971924,
+      "mean_token_accuracy": 0.8060351841151714,
+      "num_tokens": 702638.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.724200362100181,
+      "eval_entropy": 0.7380402811457601,
+      "eval_loss": 0.7054820656776428,
+      "eval_mean_token_accuracy": 0.8003534738267406,
+      "eval_num_tokens": 702638.0,
+      "eval_runtime": 88.238,
+      "eval_samples_per_second": 16.093,
+      "eval_steps_per_second": 2.017,
+      "step": 300
+    },
+    {
+      "entropy": 0.7696981988847256,
+      "epoch": 0.7724803862401931,
+      "grad_norm": 0.8408658504486084,
+      "learning_rate": 0.00028924209318837064,
+      "loss": 0.7070387363433838,
+      "mean_token_accuracy": 0.8035947173833847,
+      "num_tokens": 749377.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.7724803862401931,
+      "eval_entropy": 0.7897409023193831,
+      "eval_loss": 0.7015668749809265,
+      "eval_mean_token_accuracy": 0.805306687113944,
+      "eval_num_tokens": 749377.0,
+      "eval_runtime": 87.9395,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 320
+    },
+    {
+      "entropy": 0.762337189912796,
+      "epoch": 0.8207604103802052,
+      "grad_norm": 0.9843310117721558,
+      "learning_rate": 0.0003073763937017481,
+      "loss": 0.7021088123321533,
+      "mean_token_accuracy": 0.8029616877436638,
+      "num_tokens": 798874.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8207604103802052,
+      "eval_entropy": 0.7210077121016685,
+      "eval_loss": 0.7082746624946594,
+      "eval_mean_token_accuracy": 0.8049283824609906,
+      "eval_num_tokens": 798874.0,
+      "eval_runtime": 88.1706,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 340
+    },
+    {
+      "entropy": 0.7879025422036647,
+      "epoch": 0.8690404345202173,
+      "grad_norm": 0.9713814854621887,
+      "learning_rate": 0.0003255106942151256,
+      "loss": 0.714624547958374,
+      "mean_token_accuracy": 0.79987031519413,
+      "num_tokens": 840825.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8690404345202173,
+      "eval_entropy": 0.7866930489459735,
+      "eval_loss": 0.7145519852638245,
+      "eval_mean_token_accuracy": 0.8015657988157165,
+      "eval_num_tokens": 840825.0,
+      "eval_runtime": 88.3746,
+      "eval_samples_per_second": 16.068,
+      "eval_steps_per_second": 2.014,
+      "step": 360
+    },
+    {
+      "entropy": 0.7844726584851742,
+      "epoch": 0.9173204586602294,
+      "grad_norm": 1.2028311491012573,
+      "learning_rate": 0.00034364499472850306,
+      "loss": 0.7040786266326904,
+      "mean_token_accuracy": 0.7981885217130185,
+      "num_tokens": 882885.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9173204586602294,
+      "eval_entropy": 0.7767668339643585,
+      "eval_loss": 0.7073574066162109,
+      "eval_mean_token_accuracy": 0.8040851833445303,
+      "eval_num_tokens": 882885.0,
+      "eval_runtime": 88.3828,
+      "eval_samples_per_second": 16.066,
+      "eval_steps_per_second": 2.014,
+      "step": 380
+    },
+    {
+      "entropy": 0.7699790760874748,
+      "epoch": 0.9656004828002414,
+      "grad_norm": 0.9397398233413696,
+      "learning_rate": 0.00036177929524188054,
+      "loss": 0.7040313720703125,
+      "mean_token_accuracy": 0.8034727744758129,
+      "num_tokens": 927456.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9656004828002414,
+      "eval_entropy": 0.7749103545472863,
+      "eval_loss": 0.7093513011932373,
+      "eval_mean_token_accuracy": 0.8019482901926791,
+      "eval_num_tokens": 927456.0,
+      "eval_runtime": 86.2112,
+      "eval_samples_per_second": 16.471,
+      "eval_steps_per_second": 2.065,
+      "step": 400
+    },
+    {
+      "entropy": 0.7672389172888422,
+      "epoch": 1.012070006035003,
+      "grad_norm": 0.99210524559021,
+      "learning_rate": 0.00037628567078152296,
+      "loss": 0.6913118362426758,
+      "mean_token_accuracy": 0.8042210944287189,
+      "num_tokens": 972662.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.012070006035003,
+      "eval_entropy": 0.7294771229283193,
+      "eval_loss": 0.7171520590782166,
+      "eval_mean_token_accuracy": 0.8015128585059991,
+      "eval_num_tokens": 972662.0,
+      "eval_runtime": 88.037,
+      "eval_samples_per_second": 16.13,
+      "eval_steps_per_second": 2.022,
+      "step": 420
+    },
+    {
+      "entropy": 0.7143943183124065,
+      "epoch": 1.060350030175015,
+      "grad_norm": 1.087638258934021,
+      "learning_rate": 0.0003762484015601069,
+      "loss": 0.642704439163208,
+      "mean_token_accuracy": 0.8124966286122799,
+      "num_tokens": 1022525.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.060350030175015,
+      "eval_entropy": 0.743511886744017,
+      "eval_loss": 0.7116662859916687,
+      "eval_mean_token_accuracy": 0.8037210672758939,
+      "eval_num_tokens": 1022525.0,
+      "eval_runtime": 87.9931,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 440
+    },
+    {
+      "entropy": 0.7237346574664116,
+      "epoch": 1.1086300543150271,
+      "grad_norm": 0.9365553259849548,
+      "learning_rate": 0.0003761579008392123,
+      "loss": 0.652644681930542,
+      "mean_token_accuracy": 0.8107496216893196,
+      "num_tokens": 1069194.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1086300543150271,
+      "eval_entropy": 0.7372823839776972,
+      "eval_loss": 0.7238303422927856,
+      "eval_mean_token_accuracy": 0.8026230124275336,
+      "eval_num_tokens": 1069194.0,
+      "eval_runtime": 87.6778,
+      "eval_samples_per_second": 16.196,
+      "eval_steps_per_second": 2.03,
+      "step": 460
+    },
+    {
+      "entropy": 0.7202137842774391,
+      "epoch": 1.1569100784550392,
+      "grad_norm": 0.7998443841934204,
+      "learning_rate": 0.0003760141942294726,
+      "loss": 0.6485219478607178,
+      "mean_token_accuracy": 0.8092714451253414,
+      "num_tokens": 1117570.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1569100784550392,
+      "eval_entropy": 0.71547647942318,
+      "eval_loss": 0.7124277949333191,
+      "eval_mean_token_accuracy": 0.8013174219077892,
+      "eval_num_tokens": 1117570.0,
+      "eval_runtime": 87.4387,
+      "eval_samples_per_second": 16.24,
+      "eval_steps_per_second": 2.036,
+      "step": 480
+    },
+    {
+      "entropy": 0.7270819500088692,
+      "epoch": 1.2051901025950513,
+      "grad_norm": 0.9470787644386292,
+      "learning_rate": 0.00037581732239815854,
+      "loss": 0.66352219581604,
+      "mean_token_accuracy": 0.8102593503892421,
+      "num_tokens": 1161717.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.2051901025950513,
+      "eval_entropy": 0.7791565897759427,
+      "eval_loss": 0.7179726958274841,
+      "eval_mean_token_accuracy": 0.7948898328154275,
+      "eval_num_tokens": 1161717.0,
+      "eval_runtime": 87.9927,
+      "eval_samples_per_second": 16.138,
+      "eval_steps_per_second": 2.023,
+      "step": 500
+    },
+    {
+      "entropy": 0.7320936664938926,
+      "epoch": 1.2534701267350634,
+      "grad_norm": 1.0220379829406738,
+      "learning_rate": 0.0003755673410576695,
+      "loss": 0.6581549644470215,
+      "mean_token_accuracy": 0.8098709337413311,
+      "num_tokens": 1209390.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.2534701267350634,
+      "eval_entropy": 0.7710678225822662,
+      "eval_loss": 0.708797812461853,
+      "eval_mean_token_accuracy": 0.8039572778042783,
+      "eval_num_tokens": 1209390.0,
+      "eval_runtime": 88.1193,
+      "eval_samples_per_second": 16.115,
+      "eval_steps_per_second": 2.02,
+      "step": 520
+    },
+    {
+      "entropy": 0.7281674664467573,
+      "epoch": 1.3017501508750755,
+      "grad_norm": 0.8775188326835632,
+      "learning_rate": 0.000375264320949768,
+      "loss": 0.6631278991699219,
+      "mean_token_accuracy": 0.8101261422038079,
+      "num_tokens": 1256611.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3017501508750755,
+      "eval_entropy": 0.7199799800187014,
+      "eval_loss": 0.6981531381607056,
+      "eval_mean_token_accuracy": 0.8071265254127845,
+      "eval_num_tokens": 1256611.0,
+      "eval_runtime": 88.165,
+      "eval_samples_per_second": 16.106,
+      "eval_steps_per_second": 2.019,
+      "step": 540
+    },
+    {
+      "entropy": 0.722860773652792,
+      "epoch": 1.3500301750150876,
+      "grad_norm": 0.8949288129806519,
+      "learning_rate": 0.00037490834782556,
+      "loss": 0.6639598846435547,
+      "mean_token_accuracy": 0.8105649061501026,
+      "num_tokens": 1302325.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3500301750150876,
+      "eval_entropy": 0.7134760032209118,
+      "eval_loss": 0.7029948234558105,
+      "eval_mean_token_accuracy": 0.8062985880991046,
+      "eval_num_tokens": 1302325.0,
+      "eval_runtime": 88.1286,
+      "eval_samples_per_second": 16.113,
+      "eval_steps_per_second": 2.02,
+      "step": 560
+    },
+    {
+      "entropy": 0.7299126699566841,
+      "epoch": 1.3983101991550995,
+      "grad_norm": 1.0323039293289185,
+      "learning_rate": 0.0003744995224212291,
+      "loss": 0.6535763263702392,
+      "mean_token_accuracy": 0.8120224848389626,
+      "num_tokens": 1343844.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.3983101991550995,
+      "eval_entropy": 0.7169701182440426,
+      "eval_loss": 0.7067857980728149,
+      "eval_mean_token_accuracy": 0.8040432936689826,
+      "eval_num_tokens": 1343844.0,
+      "eval_runtime": 88.2252,
+      "eval_samples_per_second": 16.095,
+      "eval_steps_per_second": 2.018,
+      "step": 580
+    },
+    {
+      "entropy": 0.7281742256134749,
+      "epoch": 1.4465902232951118,
+      "grad_norm": 1.3370424509048462,
+      "learning_rate": 0.00037403796042952863,
+      "loss": 0.6637272357940673,
+      "mean_token_accuracy": 0.8118261776864528,
+      "num_tokens": 1389471.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4465902232951118,
+      "eval_entropy": 0.7378843212395572,
+      "eval_loss": 0.7038390040397644,
+      "eval_mean_token_accuracy": 0.8059229371922739,
+      "eval_num_tokens": 1389471.0,
+      "eval_runtime": 87.9301,
+      "eval_samples_per_second": 16.149,
+      "eval_steps_per_second": 2.024,
+      "step": 600
+    },
+    {
+      "entropy": 0.7388614989817143,
+      "epoch": 1.4948702474351236,
+      "grad_norm": 0.8464171290397644,
+      "learning_rate": 0.00037352379246704257,
+      "loss": 0.6604740142822265,
+      "mean_token_accuracy": 0.8102974124252796,
+      "num_tokens": 1435975.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.4948702474351236,
+      "eval_entropy": 0.7126847046814607,
+      "eval_loss": 0.6950703263282776,
+      "eval_mean_token_accuracy": 0.8085885392815879,
+      "eval_num_tokens": 1435975.0,
+      "eval_runtime": 88.2732,
+      "eval_samples_per_second": 16.086,
+      "eval_steps_per_second": 2.016,
+      "step": 620
+    },
+    {
+      "entropy": 0.7258916139602661,
+      "epoch": 1.5431502715751357,
+      "grad_norm": 1.1780085563659668,
+      "learning_rate": 0.0003729571640372223,
+      "loss": 0.6529891014099121,
+      "mean_token_accuracy": 0.8085981778800487,
+      "num_tokens": 1482035.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.5431502715751357,
+      "eval_entropy": 0.7096509638797032,
+      "eval_loss": 0.690051257610321,
+      "eval_mean_token_accuracy": 0.8100520241796301,
+      "eval_num_tokens": 1482035.0,
+      "eval_runtime": 88.3153,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 2.016,
+      "step": 640
+    },
+    {
+      "entropy": 0.7151173129677773,
+      "epoch": 1.5914302957151478,
+      "grad_norm": 1.0357667207717896,
+      "learning_rate": 0.0003723382354892108,
+      "loss": 0.6451474189758301,
+      "mean_token_accuracy": 0.8136902332305909,
+      "num_tokens": 1528119.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.5914302957151478,
+      "eval_entropy": 0.7060252652409371,
+      "eval_loss": 0.6899483799934387,
+      "eval_mean_token_accuracy": 0.810834194502134,
+      "eval_num_tokens": 1528119.0,
+      "eval_runtime": 87.8848,
+      "eval_samples_per_second": 16.158,
+      "eval_steps_per_second": 2.025,
+      "step": 660
+    },
+    {
+      "entropy": 0.7141377851366997,
+      "epoch": 1.63971031985516,
+      "grad_norm": 0.8391667008399963,
+      "learning_rate": 0.000371667181972466,
+      "loss": 0.6481579780578614,
+      "mean_token_accuracy": 0.8159015104174614,
+      "num_tokens": 1573661.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.63971031985516,
+      "eval_entropy": 0.7598993319473909,
+      "eval_loss": 0.681204080581665,
+      "eval_mean_token_accuracy": 0.8100194586126992,
+      "eval_num_tokens": 1573661.0,
+      "eval_runtime": 88.2415,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 680
+    },
+    {
+      "entropy": 0.7235980287194252,
+      "epoch": 1.687990343995172,
+      "grad_norm": 0.8197196125984192,
+      "learning_rate": 0.00037094419338719537,
+      "loss": 0.6582849025726318,
+      "mean_token_accuracy": 0.8114834539592266,
+      "num_tokens": 1621714.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.687990343995172,
+      "eval_entropy": 0.7166854568411795,
+      "eval_loss": 0.679177463054657,
+      "eval_mean_token_accuracy": 0.8119039877077167,
+      "eval_num_tokens": 1621714.0,
+      "eval_runtime": 88.0957,
+      "eval_samples_per_second": 16.119,
+      "eval_steps_per_second": 2.021,
+      "step": 700
+    },
+    {
+      "entropy": 0.7278190158307553,
+      "epoch": 1.736270368135184,
+      "grad_norm": 0.7820602655410767,
+      "learning_rate": 0.0003701694743306164,
+      "loss": 0.6633153438568116,
+      "mean_token_accuracy": 0.8123249113559723,
+      "num_tokens": 1672118.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.736270368135184,
+      "eval_entropy": 0.6980850314826108,
+      "eval_loss": 0.6775653958320618,
+      "eval_mean_token_accuracy": 0.8116872906684875,
+      "eval_num_tokens": 1672118.0,
+      "eval_runtime": 87.9795,
+      "eval_samples_per_second": 16.14,
+      "eval_steps_per_second": 2.023,
+      "step": 720
+    },
+    {
+      "entropy": 0.7149631556123495,
+      "epoch": 1.7845503922751962,
+      "grad_norm": 1.0139760971069336,
+      "learning_rate": 0.00036934324403905795,
+      "loss": 0.6468976020812989,
+      "mean_token_accuracy": 0.8117561548948288,
+      "num_tokens": 1720465.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.7845503922751962,
+      "eval_entropy": 0.69009849094273,
+      "eval_loss": 0.6731093525886536,
+      "eval_mean_token_accuracy": 0.81281964196248,
+      "eval_num_tokens": 1720465.0,
+      "eval_runtime": 87.7412,
+      "eval_samples_per_second": 16.184,
+      "eval_steps_per_second": 2.029,
+      "step": 740
+    },
+    {
+      "entropy": 0.7226049453020096,
+      "epoch": 1.832830416415208,
+      "grad_norm": 1.5040546655654907,
+      "learning_rate": 0.0003684657363259193,
+      "loss": 0.6385328769683838,
+      "mean_token_accuracy": 0.8132639616727829,
+      "num_tokens": 1764440.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.832830416415208,
+      "eval_entropy": 0.7125169618076153,
+      "eval_loss": 0.6864680647850037,
+      "eval_mean_token_accuracy": 0.8049418089095126,
+      "eval_num_tokens": 1764440.0,
+      "eval_runtime": 87.9429,
+      "eval_samples_per_second": 16.147,
+      "eval_steps_per_second": 2.024,
+      "step": 760
+    },
+    {
+      "entropy": 0.707821810990572,
+      "epoch": 1.8811104405552204,
+      "grad_norm": 0.8604223132133484,
+      "learning_rate": 0.00036753719951550327,
+      "loss": 0.6397049427032471,
+      "mean_token_accuracy": 0.8137877315282822,
+      "num_tokens": 1810828.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.8811104405552204,
+      "eval_entropy": 0.6492102551326323,
+      "eval_loss": 0.6743726134300232,
+      "eval_mean_token_accuracy": 0.8130356572317273,
+      "eval_num_tokens": 1810828.0,
+      "eval_runtime": 87.8684,
+      "eval_samples_per_second": 16.161,
+      "eval_steps_per_second": 2.026,
+      "step": 780
+    },
+    {
+      "entropy": 0.6800355084240437,
+      "epoch": 1.9293904646952322,
+      "grad_norm": 0.7270982265472412,
+      "learning_rate": 0.00036655789637274377,
+      "loss": 0.6267755031585693,
+      "mean_token_accuracy": 0.8180945307016373,
+      "num_tokens": 1857600.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9293904646952322,
+      "eval_entropy": 0.6727538651294922,
+      "eval_loss": 0.6696028709411621,
+      "eval_mean_token_accuracy": 0.8116443090224534,
+      "eval_num_tokens": 1857600.0,
+      "eval_runtime": 88.2449,
+      "eval_samples_per_second": 16.092,
+      "eval_steps_per_second": 2.017,
+      "step": 800
+    },
+    {
+      "entropy": 0.6746691606938839,
+      "epoch": 1.9776704888352445,
+      "grad_norm": 0.9771540760993958,
+      "learning_rate": 0.0003655281040288461,
+      "loss": 0.6326992988586426,
+      "mean_token_accuracy": 0.8179976396262646,
+      "num_tokens": 1901998.0,
+      "step": 820
+    },
+    {
+      "epoch": 1.9776704888352445,
+      "eval_entropy": 0.6636319716324967,
+      "eval_loss": 0.6714363098144531,
+      "eval_mean_token_accuracy": 0.8126635943235976,
+      "eval_num_tokens": 1901998.0,
+      "eval_runtime": 87.7666,
+      "eval_samples_per_second": 16.179,
+      "eval_steps_per_second": 2.028,
+      "step": 820
+    },
+    {
+      "entropy": 0.6319125778489298,
+      "epoch": 2.024140012070006,
+      "grad_norm": 0.8380444645881653,
+      "learning_rate": 0.0003644481139028622,
+      "loss": 0.5780903816223144,
+      "mean_token_accuracy": 0.8250188401767186,
+      "num_tokens": 1947553.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.024140012070006,
+      "eval_entropy": 0.6627151989535,
+      "eval_loss": 0.6774935722351074,
+      "eval_mean_token_accuracy": 0.8140510819601209,
+      "eval_num_tokens": 1947553.0,
+      "eval_runtime": 87.8287,
+      "eval_samples_per_second": 16.168,
+      "eval_steps_per_second": 2.027,
+      "step": 840
+    },
+    {
+      "entropy": 0.6059147048741579,
+      "epoch": 2.0724200362100182,
+      "grad_norm": 0.9745152592658997,
+      "learning_rate": 0.0003633182316192229,
+      "loss": 0.546702527999878,
+      "mean_token_accuracy": 0.8358186542987823,
+      "num_tokens": 1995234.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.0724200362100182,
+      "eval_entropy": 0.6620710384979677,
+      "eval_loss": 0.6769291162490845,
+      "eval_mean_token_accuracy": 0.8135082835561773,
+      "eval_num_tokens": 1995234.0,
+      "eval_runtime": 88.0008,
+      "eval_samples_per_second": 16.136,
+      "eval_steps_per_second": 2.023,
+      "step": 860
+    },
+    {
+      "entropy": 0.6076117843389511,
+      "epoch": 2.12070006035003,
+      "grad_norm": 0.6914380192756653,
+      "learning_rate": 0.0003621387769212491,
+      "loss": 0.5324535846710206,
+      "mean_token_accuracy": 0.8366568259894848,
+      "num_tokens": 2040639.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.12070006035003,
+      "eval_entropy": 0.6592224213514435,
+      "eval_loss": 0.670383632183075,
+      "eval_mean_token_accuracy": 0.8162150088320957,
+      "eval_num_tokens": 2040639.0,
+      "eval_runtime": 88.324,
+      "eval_samples_per_second": 16.077,
+      "eval_steps_per_second": 2.015,
+      "step": 880
+    },
+    {
+      "entropy": 0.6158072650432587,
+      "epoch": 2.1689800844900424,
+      "grad_norm": 0.9649198055267334,
+      "learning_rate": 0.0003609100835806688,
+      "loss": 0.5564568996429443,
+      "mean_token_accuracy": 0.8329654954373836,
+      "num_tokens": 2090604.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.1689800844900424,
+      "eval_entropy": 0.608588819758276,
+      "eval_loss": 0.6745102405548096,
+      "eval_mean_token_accuracy": 0.8155080542135774,
+      "eval_num_tokens": 2090604.0,
+      "eval_runtime": 87.52,
+      "eval_samples_per_second": 16.225,
+      "eval_steps_per_second": 2.034,
+      "step": 900
+    },
+    {
+      "entropy": 0.5938734326511621,
+      "epoch": 2.2172601086300543,
+      "grad_norm": 0.6915557384490967,
+      "learning_rate": 0.0003596324993031632,
+      "loss": 0.5351875782012939,
+      "mean_token_accuracy": 0.8385631121695042,
+      "num_tokens": 2142246.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2172601086300543,
+      "eval_entropy": 0.6251631205001574,
+      "eval_loss": 0.6664518117904663,
+      "eval_mean_token_accuracy": 0.81711940417129,
+      "eval_num_tokens": 2142246.0,
+      "eval_runtime": 87.1792,
+      "eval_samples_per_second": 16.288,
+      "eval_steps_per_second": 2.042,
+      "step": 920
+    },
+    {
+      "entropy": 0.623173613846302,
+      "epoch": 2.2655401327700666,
+      "grad_norm": 0.8366610407829285,
+      "learning_rate": 0.00035830638562997063,
+      "loss": 0.5488213539123535,
+      "mean_token_accuracy": 0.8321483485400677,
+      "num_tokens": 2183144.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.2655401327700666,
+      "eval_entropy": 0.6237085649136747,
+      "eval_loss": 0.68075031042099,
+      "eval_mean_token_accuracy": 0.814671738429016,
+      "eval_num_tokens": 2183144.0,
+      "eval_runtime": 88.0652,
+      "eval_samples_per_second": 16.124,
+      "eval_steps_per_second": 2.021,
+      "step": 940
+    },
+    {
+      "entropy": 0.6236911740154027,
+      "epoch": 2.3138201569100785,
+      "grad_norm": 0.7812452912330627,
+      "learning_rate": 0.00035693211783557416,
+      "loss": 0.5696200847625732,
+      "mean_token_accuracy": 0.82991396561265,
+      "num_tokens": 2228098.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3138201569100785,
+      "eval_entropy": 0.6327033426319615,
+      "eval_loss": 0.6742984652519226,
+      "eval_mean_token_accuracy": 0.8148086091105857,
+      "eval_num_tokens": 2228098.0,
+      "eval_runtime": 88.1711,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 960
+    },
+    {
+      "entropy": 0.6218426413834095,
+      "epoch": 2.3621001810500903,
+      "grad_norm": 0.8719790577888489,
+      "learning_rate": 0.0003555100848215035,
+      "loss": 0.5582651615142822,
+      "mean_token_accuracy": 0.8323961742222309,
+      "num_tokens": 2273062.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.3621001810500903,
+      "eval_entropy": 0.6338012375858393,
+      "eval_loss": 0.6685822606086731,
+      "eval_mean_token_accuracy": 0.8152180300669724,
+      "eval_num_tokens": 2273062.0,
+      "eval_runtime": 88.101,
+      "eval_samples_per_second": 16.118,
+      "eval_steps_per_second": 2.02,
+      "step": 980
+    },
+    {
+      "entropy": 0.6050681818276644,
+      "epoch": 2.4103802051901027,
+      "grad_norm": 0.9151566028594971,
+      "learning_rate": 0.00035404068900628076,
+      "loss": 0.559452486038208,
+      "mean_token_accuracy": 0.8310242861509323,
+      "num_tokens": 2319305.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.4103802051901027,
+      "eval_entropy": 0.6634890969549672,
+      "eval_loss": 0.6692465543746948,
+      "eval_mean_token_accuracy": 0.8134353673190213,
+      "eval_num_tokens": 2319305.0,
+      "eval_runtime": 87.7711,
+      "eval_samples_per_second": 16.178,
+      "eval_steps_per_second": 2.028,
+      "step": 1000
+    },
+    {
+      "entropy": 0.6169968567788601,
+      "epoch": 2.4586602293301145,
+      "grad_norm": 0.7081006765365601,
+      "learning_rate": 0.0003525243462115402,
+      "loss": 0.5589694023132324,
+      "mean_token_accuracy": 0.8357502184808254,
+      "num_tokens": 2368592.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.4586602293301145,
+      "eval_entropy": 0.6126393578695447,
+      "eval_loss": 0.6706712245941162,
+      "eval_mean_token_accuracy": 0.8155577500884452,
+      "eval_num_tokens": 2368592.0,
+      "eval_runtime": 88.1407,
+      "eval_samples_per_second": 16.111,
+      "eval_steps_per_second": 2.019,
+      "step": 1020
+    },
+    {
+      "entropy": 0.6072248566895724,
+      "epoch": 2.506940253470127,
+      "grad_norm": 0.6999370455741882,
+      "learning_rate": 0.0003509614855443561,
+      "loss": 0.5500593185424805,
+      "mean_token_accuracy": 0.8337264291942119,
+      "num_tokens": 2412859.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.506940253470127,
+      "eval_entropy": 0.6417417074187418,
+      "eval_loss": 0.6651163697242737,
+      "eval_mean_token_accuracy": 0.8159113908081912,
+      "eval_num_tokens": 2412859.0,
+      "eval_runtime": 87.4854,
+      "eval_samples_per_second": 16.231,
+      "eval_steps_per_second": 2.035,
+      "step": 1040
+    },
+    {
+      "entropy": 0.6301592070609331,
+      "epoch": 2.5552202776101387,
+      "grad_norm": 0.8768910765647888,
+      "learning_rate": 0.00034935254927581064,
+      "loss": 0.5613903999328613,
+      "mean_token_accuracy": 0.8298896946012974,
+      "num_tokens": 2458201.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.5552202776101387,
+      "eval_entropy": 0.6569794751285167,
+      "eval_loss": 0.6668341159820557,
+      "eval_mean_token_accuracy": 0.8126791795987761,
+      "eval_num_tokens": 2458201.0,
+      "eval_runtime": 88.0474,
+      "eval_samples_per_second": 16.128,
+      "eval_steps_per_second": 2.022,
+      "step": 1060
+    },
+    {
+      "entropy": 0.6069021724164486,
+      "epoch": 2.603500301750151,
+      "grad_norm": 0.9033508896827698,
+      "learning_rate": 0.0003476979927158357,
+      "loss": 0.5590654373168945,
+      "mean_token_accuracy": 0.831520090252161,
+      "num_tokens": 2505382.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.603500301750151,
+      "eval_entropy": 0.6636380777600106,
+      "eval_loss": 0.6596238017082214,
+      "eval_mean_token_accuracy": 0.8177202826135614,
+      "eval_num_tokens": 2505382.0,
+      "eval_runtime": 88.08,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1080
+    },
+    {
+      "entropy": 0.6138021353632211,
+      "epoch": 2.651780325890163,
+      "grad_norm": 0.7345808148384094,
+      "learning_rate": 0.0003459982840843664,
+      "loss": 0.5586410522460937,
+      "mean_token_accuracy": 0.8295876093208789,
+      "num_tokens": 2552927.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.651780325890163,
+      "eval_entropy": 0.5845904127600487,
+      "eval_loss": 0.663780689239502,
+      "eval_mean_token_accuracy": 0.8142095107710763,
+      "eval_num_tokens": 2552927.0,
+      "eval_runtime": 88.0083,
+      "eval_samples_per_second": 16.135,
+      "eval_steps_per_second": 2.023,
+      "step": 1100
+    },
+    {
+      "entropy": 0.6251844003796577,
+      "epoch": 2.700060350030175,
+      "grad_norm": 0.7587347626686096,
+      "learning_rate": 0.00034425390437883976,
+      "loss": 0.5720802307128906,
+      "mean_token_accuracy": 0.8285771444439888,
+      "num_tokens": 2597426.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.700060350030175,
+      "eval_entropy": 0.6251207221759839,
+      "eval_loss": 0.6556326746940613,
+      "eval_mean_token_accuracy": 0.8181774036937886,
+      "eval_num_tokens": 2597426.0,
+      "eval_runtime": 87.8517,
+      "eval_samples_per_second": 16.164,
+      "eval_steps_per_second": 2.026,
+      "step": 1120
+    },
+    {
+      "entropy": 0.6227757651358843,
+      "epoch": 2.748340374170187,
+      "grad_norm": 0.9254975914955139,
+      "learning_rate": 0.00034246534723807843,
+      "loss": 0.5767871856689453,
+      "mean_token_accuracy": 0.8292960874736309,
+      "num_tokens": 2640514.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.748340374170187,
+      "eval_entropy": 0.6580858331048087,
+      "eval_loss": 0.6579018831253052,
+      "eval_mean_token_accuracy": 0.8178701273510965,
+      "eval_num_tokens": 2640514.0,
+      "eval_runtime": 87.9523,
+      "eval_samples_per_second": 16.145,
+      "eval_steps_per_second": 2.024,
+      "step": 1140
+    },
+    {
+      "entropy": 0.6353523321449757,
+      "epoch": 2.796620398310199,
+      "grad_norm": 1.4437663555145264,
+      "learning_rate": 0.00034063311880259663,
+      "loss": 0.566702127456665,
+      "mean_token_accuracy": 0.831210870295763,
+      "num_tokens": 2687730.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.796620398310199,
+      "eval_entropy": 0.6347215423423253,
+      "eval_loss": 0.6531020402908325,
+      "eval_mean_token_accuracy": 0.8189373260803436,
+      "eval_num_tokens": 2687730.0,
+      "eval_runtime": 88.0207,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1160
+    },
+    {
+      "entropy": 0.6233414955437183,
+      "epoch": 2.8449004224502112,
+      "grad_norm": 0.753302812576294,
+      "learning_rate": 0.0003387577375713684,
+      "loss": 0.5740012645721435,
+      "mean_token_accuracy": 0.8270302519202233,
+      "num_tokens": 2732930.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.8449004224502112,
+      "eval_entropy": 0.6291426423560368,
+      "eval_loss": 0.6549608707427979,
+      "eval_mean_token_accuracy": 0.8188753693961026,
+      "eval_num_tokens": 2732930.0,
+      "eval_runtime": 88.2306,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1180
+    },
+    {
+      "entropy": 0.617981968075037,
+      "epoch": 2.8931804465902236,
+      "grad_norm": 0.6777492165565491,
+      "learning_rate": 0.00033683973425509874,
+      "loss": 0.5558822631835938,
+      "mean_token_accuracy": 0.8315935507416725,
+      "num_tokens": 2779450.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.8931804465902236,
+      "eval_entropy": 0.6340839514571629,
+      "eval_loss": 0.6490646600723267,
+      "eval_mean_token_accuracy": 0.8200173856837026,
+      "eval_num_tokens": 2779450.0,
+      "eval_runtime": 88.3076,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1200
+    },
+    {
+      "entropy": 0.5989151727408171,
+      "epoch": 2.9414604707302354,
+      "grad_norm": 1.0503766536712646,
+      "learning_rate": 0.0003348796516260381,
+      "loss": 0.550087308883667,
+      "mean_token_accuracy": 0.8357349015772343,
+      "num_tokens": 2827819.0,
+      "step": 1220
+    },
+    {
+      "epoch": 2.9414604707302354,
+      "eval_entropy": 0.6683844726407127,
+      "eval_loss": 0.6455520987510681,
+      "eval_mean_token_accuracy": 0.8199967946229356,
+      "eval_num_tokens": 2827819.0,
+      "eval_runtime": 87.7825,
+      "eval_samples_per_second": 16.176,
+      "eval_steps_per_second": 2.028,
+      "step": 1220
+    },
+    {
+      "entropy": 0.6396764755249024,
+      "epoch": 2.9897404948702473,
+      "grad_norm": 0.7088311910629272,
+      "learning_rate": 0.00033287804436438515,
+      "loss": 0.5731344223022461,
+      "mean_token_accuracy": 0.8298504173755645,
+      "num_tokens": 2872493.0,
+      "step": 1240
+    },
+    {
+      "epoch": 2.9897404948702473,
+      "eval_entropy": 0.6510144954317072,
+      "eval_loss": 0.6509669423103333,
+      "eval_mean_token_accuracy": 0.8188618431600292,
+      "eval_num_tokens": 2872493.0,
+      "eval_runtime": 87.8954,
+      "eval_samples_per_second": 16.156,
+      "eval_steps_per_second": 2.025,
+      "step": 1240
+    },
+    {
+      "entropy": 0.5436535887903982,
+      "epoch": 3.036210018105009,
+      "grad_norm": 0.8652629852294922,
+      "learning_rate": 0.00033083547890131815,
+      "loss": 0.4816119194030762,
+      "mean_token_accuracy": 0.8523989869402601,
+      "num_tokens": 2918249.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.036210018105009,
+      "eval_entropy": 0.5676146091035242,
+      "eval_loss": 0.6723794341087341,
+      "eval_mean_token_accuracy": 0.8197960103495737,
+      "eval_num_tokens": 2918249.0,
+      "eval_runtime": 88.1303,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1260
+    },
+    {
+      "entropy": 0.503798121958971,
+      "epoch": 3.084490042245021,
+      "grad_norm": 0.9403247237205505,
+      "learning_rate": 0.0003287525332587019,
+      "loss": 0.43859500885009767,
+      "mean_token_accuracy": 0.8604862734675407,
+      "num_tokens": 2968353.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.084490042245021,
+      "eval_entropy": 0.5981672088081917,
+      "eval_loss": 0.671315610408783,
+      "eval_mean_token_accuracy": 0.8175233046660263,
+      "eval_num_tokens": 2968353.0,
+      "eval_runtime": 88.3097,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1280
+    },
+    {
+      "entropy": 0.5218576818704606,
+      "epoch": 3.1327700663850333,
+      "grad_norm": 1.0271061658859253,
+      "learning_rate": 0.0003266297968855144,
+      "loss": 0.45096940994262696,
+      "mean_token_accuracy": 0.8555397011339665,
+      "num_tokens": 3018291.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.1327700663850333,
+      "eval_entropy": 0.5708769342202819,
+      "eval_loss": 0.6631077527999878,
+      "eval_mean_token_accuracy": 0.8207426647121987,
+      "eval_num_tokens": 3018291.0,
+      "eval_runtime": 87.7272,
+      "eval_samples_per_second": 16.187,
+      "eval_steps_per_second": 2.029,
+      "step": 1300
+    },
+    {
+      "entropy": 0.5191332377493382,
+      "epoch": 3.181050090525045,
+      "grad_norm": 0.958055317401886,
+      "learning_rate": 0.0003244678704910397,
+      "loss": 0.45679512023925783,
+      "mean_token_accuracy": 0.8554921194911003,
+      "num_tokens": 3070712.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.181050090525045,
+      "eval_entropy": 0.5703962919082535,
+      "eval_loss": 0.6597784161567688,
+      "eval_mean_token_accuracy": 0.821146364962117,
+      "eval_num_tokens": 3070712.0,
+      "eval_runtime": 88.4154,
+      "eval_samples_per_second": 16.061,
+      "eval_steps_per_second": 2.013,
+      "step": 1320
+    },
+    {
+      "entropy": 0.5207117382436991,
+      "epoch": 3.2293301146650575,
+      "grad_norm": 0.7903274297714233,
+      "learning_rate": 0.00032226736587487434,
+      "loss": 0.44762430191040037,
+      "mean_token_accuracy": 0.8578196220099926,
+      "num_tokens": 3116564.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.2293301146650575,
+      "eval_entropy": 0.5453257858753204,
+      "eval_loss": 0.6688455939292908,
+      "eval_mean_token_accuracy": 0.8215647212574991,
+      "eval_num_tokens": 3116564.0,
+      "eval_runtime": 88.0214,
+      "eval_samples_per_second": 16.132,
+      "eval_steps_per_second": 2.022,
+      "step": 1340
+    },
+    {
+      "entropy": 0.5263226825743914,
+      "epoch": 3.2776101388050694,
+      "grad_norm": 0.8782205581665039,
+      "learning_rate": 0.00032002890575379524,
+      "loss": 0.46592864990234373,
+      "mean_token_accuracy": 0.8551752805709839,
+      "num_tokens": 3164054.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.2776101388050694,
+      "eval_entropy": 0.5805480952343244,
+      "eval_loss": 0.6707106232643127,
+      "eval_mean_token_accuracy": 0.8191288441754458,
+      "eval_num_tokens": 3164054.0,
+      "eval_runtime": 88.2293,
+      "eval_samples_per_second": 16.094,
+      "eval_steps_per_second": 2.017,
+      "step": 1360
+    },
+    {
+      "entropy": 0.5454599127173424,
+      "epoch": 3.3258901629450817,
+      "grad_norm": 0.9048764705657959,
+      "learning_rate": 0.0003177531235855382,
+      "loss": 0.4763782501220703,
+      "mean_token_accuracy": 0.8505888104438781,
+      "num_tokens": 3207419.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.3258901629450817,
+      "eval_entropy": 0.5952914317002457,
+      "eval_loss": 0.662697970867157,
+      "eval_mean_token_accuracy": 0.819491108146946,
+      "eval_num_tokens": 3207419.0,
+      "eval_runtime": 87.9869,
+      "eval_samples_per_second": 16.139,
+      "eval_steps_per_second": 2.023,
+      "step": 1380
+    },
+    {
+      "entropy": 0.5299010306596756,
+      "epoch": 3.3741701870850935,
+      "grad_norm": 0.8653950691223145,
+      "learning_rate": 0.000315440663389537,
+      "loss": 0.4581890106201172,
+      "mean_token_accuracy": 0.8536585591733455,
+      "num_tokens": 3252679.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.3741701870850935,
+      "eval_entropy": 0.5326506612675913,
+      "eval_loss": 0.6679039001464844,
+      "eval_mean_token_accuracy": 0.8209870109397374,
+      "eval_num_tokens": 3252679.0,
+      "eval_runtime": 88.3059,
+      "eval_samples_per_second": 16.08,
+      "eval_steps_per_second": 2.016,
+      "step": 1400
+    },
+    {
+      "entropy": 0.5074611511081457,
+      "epoch": 3.4224502112251054,
+      "grad_norm": 0.7786624431610107,
+      "learning_rate": 0.00031309217956467325,
+      "loss": 0.45597033500671386,
+      "mean_token_accuracy": 0.856324628740549,
+      "num_tokens": 3297309.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.4224502112251054,
+      "eval_entropy": 0.5424539073464576,
+      "eval_loss": 0.6710956692695618,
+      "eval_mean_token_accuracy": 0.8209191105338964,
+      "eval_num_tokens": 3297309.0,
+      "eval_runtime": 87.2334,
+      "eval_samples_per_second": 16.278,
+      "eval_steps_per_second": 2.041,
+      "step": 1420
+    },
+    {
+      "entropy": 0.5192026473581791,
+      "epoch": 3.4707302353651177,
+      "grad_norm": 0.7682847380638123,
+      "learning_rate": 0.0003107083367040895,
+      "loss": 0.46319098472595216,
+      "mean_token_accuracy": 0.8517456069588661,
+      "num_tokens": 3341825.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.4707302353651177,
+      "eval_entropy": 0.569345246372598,
+      "eval_loss": 0.6640089154243469,
+      "eval_mean_token_accuracy": 0.8204013628236363,
+      "eval_num_tokens": 3341825.0,
+      "eval_runtime": 87.668,
+      "eval_samples_per_second": 16.197,
+      "eval_steps_per_second": 2.03,
+      "step": 1440
+    },
+    {
+      "entropy": 0.5302285175770521,
+      "epoch": 3.51901025950513,
+      "grad_norm": 0.9500018954277039,
+      "learning_rate": 0.00030828980940711723,
+      "loss": 0.4712235927581787,
+      "mean_token_accuracy": 0.8510453663766384,
+      "num_tokens": 3388463.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.51901025950513,
+      "eval_entropy": 0.5648383499866121,
+      "eval_loss": 0.6629963517189026,
+      "eval_mean_token_accuracy": 0.8197558281126986,
+      "eval_num_tokens": 3388463.0,
+      "eval_runtime": 87.9214,
+      "eval_samples_per_second": 16.151,
+      "eval_steps_per_second": 2.025,
+      "step": 1460
+    },
+    {
+      "entropy": 0.5347354769706726,
+      "epoch": 3.567290283645142,
+      "grad_norm": 0.9935700297355652,
+      "learning_rate": 0.00030583728208837305,
+      "loss": 0.47901196479797364,
+      "mean_token_accuracy": 0.8491650126874447,
+      "num_tokens": 3433731.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.567290283645142,
+      "eval_entropy": 0.5595698517360045,
+      "eval_loss": 0.6594937443733215,
+      "eval_mean_token_accuracy": 0.8217685289597243,
+      "eval_num_tokens": 3433731.0,
+      "eval_runtime": 88.1348,
+      "eval_samples_per_second": 16.112,
+      "eval_steps_per_second": 2.02,
+      "step": 1480
+    },
+    {
+      "entropy": 0.5250448524951935,
+      "epoch": 3.6155703077851538,
+      "grad_norm": 0.8686122894287109,
+      "learning_rate": 0.0003033514487840775,
+      "loss": 0.48284616470336916,
+      "mean_token_accuracy": 0.8505332618951797,
+      "num_tokens": 3478600.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.6155703077851538,
+      "eval_entropy": 0.5399963330017047,
+      "eval_loss": 0.6682186722755432,
+      "eval_mean_token_accuracy": 0.8201540199558387,
+      "eval_num_tokens": 3478600.0,
+      "eval_runtime": 87.9383,
+      "eval_samples_per_second": 16.148,
+      "eval_steps_per_second": 2.024,
+      "step": 1500
+    },
+    {
+      "entropy": 0.5496356297284365,
+      "epoch": 3.663850331925166,
+      "grad_norm": 0.9891234636306763,
+      "learning_rate": 0.00030083301295565117,
+      "loss": 0.4842066287994385,
+      "mean_token_accuracy": 0.8493696190416813,
+      "num_tokens": 3521787.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.663850331925166,
+      "eval_entropy": 0.5889502502559276,
+      "eval_loss": 0.6597762703895569,
+      "eval_mean_token_accuracy": 0.819690072469497,
+      "eval_num_tokens": 3521787.0,
+      "eval_runtime": 88.0165,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1520
+    },
+    {
+      "entropy": 0.5289484921842813,
+      "epoch": 3.712130356065178,
+      "grad_norm": 0.7241885662078857,
+      "learning_rate": 0.00029828268729064326,
+      "loss": 0.48619561195373534,
+      "mean_token_accuracy": 0.8518108949065208,
+      "num_tokens": 3569036.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.712130356065178,
+      "eval_entropy": 0.5958652837892596,
+      "eval_loss": 0.6565147042274475,
+      "eval_mean_token_accuracy": 0.8217123010185328,
+      "eval_num_tokens": 3569036.0,
+      "eval_runtime": 88.0126,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1540
+    },
+    {
+      "entropy": 0.5450621031224727,
+      "epoch": 3.7604103802051903,
+      "grad_norm": 0.7989155650138855,
+      "learning_rate": 0.0002957011935010495,
+      "loss": 0.48642635345458984,
+      "mean_token_accuracy": 0.8462958924472332,
+      "num_tokens": 3614313.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.7604103802051903,
+      "eval_entropy": 0.5554354205895006,
+      "eval_loss": 0.6484317183494568,
+      "eval_mean_token_accuracy": 0.823845865016573,
+      "eval_num_tokens": 3614313.0,
+      "eval_runtime": 88.2711,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1560
+    },
+    {
+      "entropy": 0.5318546604365111,
+      "epoch": 3.808690404345202,
+      "grad_norm": 1.0375983715057373,
+      "learning_rate": 0.00029308926211907624,
+      "loss": 0.4860549449920654,
+      "mean_token_accuracy": 0.8488477177917957,
+      "num_tokens": 3659786.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.808690404345202,
+      "eval_entropy": 0.5575427207049359,
+      "eval_loss": 0.659203827381134,
+      "eval_mean_token_accuracy": 0.8210098288032446,
+      "eval_num_tokens": 3659786.0,
+      "eval_runtime": 88.3024,
+      "eval_samples_per_second": 16.081,
+      "eval_steps_per_second": 2.016,
+      "step": 1580
+    },
+    {
+      "entropy": 0.5312924966216087,
+      "epoch": 3.856970428485214,
+      "grad_norm": 0.9233641624450684,
+      "learning_rate": 0.00029044763229040845,
+      "loss": 0.48835930824279783,
+      "mean_token_accuracy": 0.8447530090808868,
+      "num_tokens": 3706362.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.856970428485214,
+      "eval_entropy": 0.5433843780434533,
+      "eval_loss": 0.6514624953269958,
+      "eval_mean_token_accuracy": 0.8230627580974879,
+      "eval_num_tokens": 3706362.0,
+      "eval_runtime": 87.9727,
+      "eval_samples_per_second": 16.141,
+      "eval_steps_per_second": 2.023,
+      "step": 1600
+    },
+    {
+      "entropy": 0.5400687273591757,
+      "epoch": 3.9052504526252263,
+      "grad_norm": 0.7173060178756714,
+      "learning_rate": 0.00028777705156504007,
+      "loss": 0.486147403717041,
+      "mean_token_accuracy": 0.8478169530630112,
+      "num_tokens": 3753013.0,
+      "step": 1620
+    },
+    {
+      "epoch": 3.9052504526252263,
+      "eval_entropy": 0.5856067625324378,
+      "eval_loss": 0.6469284892082214,
+      "eval_mean_token_accuracy": 0.8219244604030352,
+      "eval_num_tokens": 3753013.0,
+      "eval_runtime": 88.0776,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1620
+    },
+    {
+      "entropy": 0.5115721028298139,
+      "epoch": 3.9535304767652386,
+      "grad_norm": 1.0881255865097046,
+      "learning_rate": 0.00028507827568572655,
+      "loss": 0.47065014839172364,
+      "mean_token_accuracy": 0.8535217240452766,
+      "num_tokens": 3800024.0,
+      "step": 1640
+    },
+    {
+      "epoch": 3.9535304767652386,
+      "eval_entropy": 0.565960791003838,
+      "eval_loss": 0.6459378004074097,
+      "eval_mean_token_accuracy": 0.8246003888296277,
+      "eval_num_tokens": 3800024.0,
+      "eval_runtime": 88.4974,
+      "eval_samples_per_second": 16.046,
+      "eval_steps_per_second": 2.011,
+      "step": 1640
+    },
+    {
+      "entropy": 0.5376694152107486,
+      "epoch": 4.0,
+      "grad_norm": 3.4068243503570557,
+      "learning_rate": 0.0002823520683741183,
+      "loss": 0.4839733600616455,
+      "mean_token_accuracy": 0.8507238850965129,
+      "num_tokens": 3844304.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.0,
+      "eval_entropy": 0.5631327726198047,
+      "eval_loss": 0.6423600316047668,
+      "eval_mean_token_accuracy": 0.82502972175566,
+      "eval_num_tokens": 3844304.0,
+      "eval_runtime": 88.1524,
+      "eval_samples_per_second": 16.108,
+      "eval_steps_per_second": 2.019,
+      "step": 1660
+    },
+    {
+      "entropy": 0.4188325941562653,
+      "epoch": 4.048280024140012,
+      "grad_norm": 0.6339647173881531,
+      "learning_rate": 0.00027959920111463656,
+      "loss": 0.3500279426574707,
+      "mean_token_accuracy": 0.8840543255209923,
+      "num_tokens": 3890738.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.048280024140012,
+      "eval_entropy": 0.5159014272555876,
+      "eval_loss": 0.6822672486305237,
+      "eval_mean_token_accuracy": 0.8223584498582261,
+      "eval_num_tokens": 3890738.0,
+      "eval_runtime": 87.8011,
+      "eval_samples_per_second": 16.173,
+      "eval_steps_per_second": 2.027,
+      "step": 1680
+    },
+    {
+      "entropy": 0.42600946351885793,
+      "epoch": 4.096560048280024,
+      "grad_norm": 0.6938858032226562,
+      "learning_rate": 0.0002768204529361524,
+      "loss": 0.3462975263595581,
+      "mean_token_accuracy": 0.8850771695375442,
+      "num_tokens": 3938205.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.096560048280024,
+      "eval_entropy": 0.4818309486916896,
+      "eval_loss": 0.697238028049469,
+      "eval_mean_token_accuracy": 0.822570590825563,
+      "eval_num_tokens": 3938205.0,
+      "eval_runtime": 87.6294,
+      "eval_samples_per_second": 16.205,
+      "eval_steps_per_second": 2.031,
+      "step": 1700
+    },
+    {
+      "entropy": 0.4128956265747547,
+      "epoch": 4.1448400724200365,
+      "grad_norm": 0.7023837566375732,
+      "learning_rate": 0.000274016610191531,
+      "loss": 0.35343332290649415,
+      "mean_token_accuracy": 0.8837565891444683,
+      "num_tokens": 3986350.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.1448400724200365,
+      "eval_entropy": 0.4736787666765492,
+      "eval_loss": 0.686392605304718,
+      "eval_mean_token_accuracy": 0.8246323891570059,
+      "eval_num_tokens": 3986350.0,
+      "eval_runtime": 87.755,
+      "eval_samples_per_second": 16.181,
+      "eval_steps_per_second": 2.028,
+      "step": 1720
+    },
+    {
+      "entropy": 0.4070820797234774,
+      "epoch": 4.193120096560048,
+      "grad_norm": 0.7664099335670471,
+      "learning_rate": 0.00027118846633510276,
+      "loss": 0.35689282417297363,
+      "mean_token_accuracy": 0.8821764968335628,
+      "num_tokens": 4033541.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.193120096560048,
+      "eval_entropy": 0.48123268063148755,
+      "eval_loss": 0.6800221800804138,
+      "eval_mean_token_accuracy": 0.8234515280536051,
+      "eval_num_tokens": 4033541.0,
+      "eval_runtime": 87.613,
+      "eval_samples_per_second": 16.208,
+      "eval_steps_per_second": 2.032,
+      "step": 1740
+    },
+    {
+      "entropy": 0.4199690703302622,
+      "epoch": 4.24140012070006,
+      "grad_norm": 0.7411720752716064,
+      "learning_rate": 0.00026833682169812544,
+      "loss": 0.35881006717681885,
+      "mean_token_accuracy": 0.8812312394380569,
+      "num_tokens": 4079253.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.24140012070006,
+      "eval_entropy": 0.47287314242861245,
+      "eval_loss": 0.691737949848175,
+      "eval_mean_token_accuracy": 0.8229516232281588,
+      "eval_num_tokens": 4079253.0,
+      "eval_runtime": 87.9596,
+      "eval_samples_per_second": 16.144,
+      "eval_steps_per_second": 2.024,
+      "step": 1760
+    },
+    {
+      "entropy": 0.4195733394473791,
+      "epoch": 4.289680144840072,
+      "grad_norm": 0.9975268244743347,
+      "learning_rate": 0.00026546248326229954,
+      "loss": 0.36998801231384276,
+      "mean_token_accuracy": 0.8777976937592029,
+      "num_tokens": 4125495.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.289680144840072,
+      "eval_entropy": 0.48877454825331657,
+      "eval_loss": 0.6843481063842773,
+      "eval_mean_token_accuracy": 0.8226028995567494,
+      "eval_num_tokens": 4125495.0,
+      "eval_runtime": 88.169,
+      "eval_samples_per_second": 16.105,
+      "eval_steps_per_second": 2.019,
+      "step": 1780
+    },
+    {
+      "entropy": 0.4087799321860075,
+      "epoch": 4.337960168980085,
+      "grad_norm": 0.8101420402526855,
+      "learning_rate": 0.00026256626443140217,
+      "loss": 0.3555623531341553,
+      "mean_token_accuracy": 0.8815289497375488,
+      "num_tokens": 4173422.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.337960168980085,
+      "eval_entropy": 0.4845838766084628,
+      "eval_loss": 0.6782916188240051,
+      "eval_mean_token_accuracy": 0.8235659428526846,
+      "eval_num_tokens": 4173422.0,
+      "eval_runtime": 88.0763,
+      "eval_samples_per_second": 16.122,
+      "eval_steps_per_second": 2.021,
+      "step": 1800
+    },
+    {
+      "entropy": 0.4086425334215164,
+      "epoch": 4.386240193120097,
+      "grad_norm": 0.8936498761177063,
+      "learning_rate": 0.0002596489848011036,
+      "loss": 0.3710860013961792,
+      "mean_token_accuracy": 0.8773891530930996,
+      "num_tokens": 4216243.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.386240193120097,
+      "eval_entropy": 0.49018637685293565,
+      "eval_loss": 0.6762362718582153,
+      "eval_mean_token_accuracy": 0.8242661363623115,
+      "eval_num_tokens": 4216243.0,
+      "eval_runtime": 88.3684,
+      "eval_samples_per_second": 16.069,
+      "eval_steps_per_second": 2.014,
+      "step": 1820
+    },
+    {
+      "entropy": 0.41956189945340155,
+      "epoch": 4.434520217260109,
+      "grad_norm": 0.6450079083442688,
+      "learning_rate": 0.00025671146992703106,
+      "loss": 0.3715682029724121,
+      "mean_token_accuracy": 0.878043319284916,
+      "num_tokens": 4262463.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.434520217260109,
+      "eval_entropy": 0.4821142643020394,
+      "eval_loss": 0.6749269962310791,
+      "eval_mean_token_accuracy": 0.8252766048640348,
+      "eval_num_tokens": 4262463.0,
+      "eval_runtime": 87.6946,
+      "eval_samples_per_second": 16.193,
+      "eval_steps_per_second": 2.03,
+      "step": 1840
+    },
+    {
+      "entropy": 0.40638086050748823,
+      "epoch": 4.4828002414001205,
+      "grad_norm": 0.8167530298233032,
+      "learning_rate": 0.0002537545510911465,
+      "loss": 0.3652146100997925,
+      "mean_token_accuracy": 0.8761600315570831,
+      "num_tokens": 4311451.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.4828002414001205,
+      "eval_entropy": 0.5017492558514134,
+      "eval_loss": 0.6696512699127197,
+      "eval_mean_token_accuracy": 0.8252955695216575,
+      "eval_num_tokens": 4311451.0,
+      "eval_runtime": 88.02,
+      "eval_samples_per_second": 16.133,
+      "eval_steps_per_second": 2.022,
+      "step": 1860
+    },
+    {
+      "entropy": 0.41664722077548505,
+      "epoch": 4.531080265540133,
+      "grad_norm": 0.757738471031189,
+      "learning_rate": 0.0002507790650665036,
+      "loss": 0.368887734413147,
+      "mean_token_accuracy": 0.8786893397569656,
+      "num_tokens": 4357724.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.531080265540133,
+      "eval_entropy": 0.4798511176296834,
+      "eval_loss": 0.6738249659538269,
+      "eval_mean_token_accuracy": 0.8250340461061242,
+      "eval_num_tokens": 4357724.0,
+      "eval_runtime": 88.2713,
+      "eval_samples_per_second": 16.087,
+      "eval_steps_per_second": 2.017,
+      "step": 1880
+    },
+    {
+      "entropy": 0.42732664197683334,
+      "epoch": 4.579360289680145,
+      "grad_norm": 0.7890474200248718,
+      "learning_rate": 0.00024778585388045044,
+      "loss": 0.38672595024108886,
+      "mean_token_accuracy": 0.8742976360023021,
+      "num_tokens": 4401934.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.579360289680145,
+      "eval_entropy": 0.49321879295820603,
+      "eval_loss": 0.6752269268035889,
+      "eval_mean_token_accuracy": 0.824115453811174,
+      "eval_num_tokens": 4401934.0,
+      "eval_runtime": 87.3378,
+      "eval_samples_per_second": 16.259,
+      "eval_steps_per_second": 2.038,
+      "step": 1900
+    },
+    {
+      "entropy": 0.41300065964460375,
+      "epoch": 4.627640313820157,
+      "grad_norm": 0.866765022277832,
+      "learning_rate": 0.00024477576457634665,
+      "loss": 0.364247989654541,
+      "mean_token_accuracy": 0.8774793080985546,
+      "num_tokens": 4449745.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.627640313820157,
+      "eval_entropy": 0.49179744352115673,
+      "eval_loss": 0.6731056571006775,
+      "eval_mean_token_accuracy": 0.8245729244826885,
+      "eval_num_tokens": 4449745.0,
+      "eval_runtime": 88.2458,
+      "eval_samples_per_second": 16.091,
+      "eval_steps_per_second": 2.017,
+      "step": 1920
+    },
+    {
+      "entropy": 0.42646181732416155,
+      "epoch": 4.675920337960169,
+      "grad_norm": 0.7921772003173828,
+      "learning_rate": 0.00024174964897385928,
+      "loss": 0.3822937488555908,
+      "mean_token_accuracy": 0.8755100265145301,
+      "num_tokens": 4495074.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.675920337960169,
+      "eval_entropy": 0.5194089940089858,
+      "eval_loss": 0.6604889631271362,
+      "eval_mean_token_accuracy": 0.8260471847619903,
+      "eval_num_tokens": 4495074.0,
+      "eval_runtime": 88.0139,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.022,
+      "step": 1940
+    },
+    {
+      "entropy": 0.4235709119588137,
+      "epoch": 4.724200362100181,
+      "grad_norm": 0.8816011548042297,
+      "learning_rate": 0.00023870836342790837,
+      "loss": 0.3772272109985352,
+      "mean_token_accuracy": 0.8731484808027744,
+      "num_tokens": 4541053.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.724200362100181,
+      "eval_entropy": 0.47387497545628066,
+      "eval_loss": 0.6644074320793152,
+      "eval_mean_token_accuracy": 0.8271099258674665,
+      "eval_num_tokens": 4541053.0,
+      "eval_runtime": 88.393,
+      "eval_samples_per_second": 16.065,
+      "eval_steps_per_second": 2.014,
+      "step": 1960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4150,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.91863931572906e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f3131f7a59f38140099b6f22f325aeb538af96d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-20/trainer_state.json
@@ -0,0 +1,55 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.049813200498132,
+  "eval_steps": 20,
+  "global_step": 20,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2024853289033728.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..31b4a66609035bd5da49e13e8a4b203ac981dd9a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-200/trainer_state.json
@@ -0,0 +1,244 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.49813200498132004,
+  "eval_steps": 20,
+  "global_step": 200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.0027099623612416e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a5e24c23e7414eb8e07c96fafe2bc2ac51041c5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2000/trainer_state.json
@@ -0,0 +1,2134 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.976338729763388,
+  "eval_steps": 20,
+  "global_step": 2000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.973209163101225e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d84d74180b651ecd849288aa294a13fe367aaf48
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2020/trainer_state.json
@@ -0,0 +1,2155 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.024906600249066,
+  "eval_steps": 20,
+  "global_step": 2020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.9920107951972352e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d748ba373d102b22b9484ae81d325898095e579b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2040/trainer_state.json
@@ -0,0 +1,2176 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.074719800747198,
+  "eval_steps": 20,
+  "global_step": 2040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.0125268072666726e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f4ded21a4214e4f0d66ecad689776647bc88899
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2060/trainer_state.json
@@ -0,0 +1,2197 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.12453300124533,
+  "eval_steps": 20,
+  "global_step": 2060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.0320025227972608e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8816c25ba75f18f339c30bc6541fc2b80070b47
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2080/trainer_state.json
@@ -0,0 +1,2218 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.174346201743462,
+  "eval_steps": 20,
+  "global_step": 2080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.052914368148521e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a341474e032c17badf23a326271e95831a2a75c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2100/trainer_state.json
@@ -0,0 +1,2239 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.224159402241594,
+  "eval_steps": 20,
+  "global_step": 2100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.0744841404738765e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..22b53cd2a737c34fdb31b1b9d856e72a241265c9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2120/trainer_state.json
@@ -0,0 +1,2260 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.273972602739726,
+  "eval_steps": 20,
+  "global_step": 2120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.0924151188661043e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..18b1bd90edc78e283a65985979cab9cd062893ed
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2140/trainer_state.json
@@ -0,0 +1,2281 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.323785803237858,
+  "eval_steps": 20,
+  "global_step": 2140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.111302021170217e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a00e12d343a7264ccbc8987eb10dd290e0a41432
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2160/trainer_state.json
@@ -0,0 +1,2302 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.37359900373599,
+  "eval_steps": 20,
+  "global_step": 2160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.1310784263820083e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..07f9e15090fdf880338835d2ee6ab951ef9a30cf
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2180/trainer_state.json
@@ -0,0 +1,2323 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.423412204234122,
+  "eval_steps": 20,
+  "global_step": 2180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.1495650074986086e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7bd9a34976e5920edaf7e69350ec73402fed26f9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-220/trainer_state.json
@@ -0,0 +1,265 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.547945205479452,
+  "eval_steps": 20,
+  "global_step": 220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.2106705362059264e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6910bf18f6872fcfabe995e78d90a96cf3f3392f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2200/trainer_state.json
@@ -0,0 +1,2344 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.473225404732254,
+  "eval_steps": 20,
+  "global_step": 2200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.1719847891615744e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..acb11849d9ad494a177d4c3e7d1213c6f72dbbc6
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2220/trainer_state.json
@@ -0,0 +1,2365 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.523038605230386,
+  "eval_steps": 20,
+  "global_step": 2220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.190705638955172e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4f5c355b98c44af536eee754055e21173d274c9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2240/trainer_state.json
@@ -0,0 +1,2386 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.572851805728518,
+  "eval_steps": 20,
+  "global_step": 2240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.2098716889931776e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff9b0e7e556216ccf17bfb7f6584f5f678630bd7
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2260/trainer_state.json
@@ -0,0 +1,2407 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.62266500622665,
+  "eval_steps": 20,
+  "global_step": 2260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.2298195322023117e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d59f7088467b83b3d0982d3bc1ff038c5e69eedc
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2280/trainer_state.json
@@ -0,0 +1,2428 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.672478206724782,
+  "eval_steps": 20,
+  "global_step": 2280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.2491624057244877e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..937774316cec4875988dcc1b4d47bddcb6f98a19
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2300/trainer_state.json
@@ -0,0 +1,2449 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.722291407222914,
+  "eval_steps": 20,
+  "global_step": 2300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.268772758425764e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..97742f684137dd910948fc680fc35337c185ff0a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2320/trainer_state.json
@@ -0,0 +1,2470 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.772104607721046,
+  "eval_steps": 20,
+  "global_step": 2320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.288900117862482e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd1a763b4d3a290a998d420d57f4f1afaeb3ebfe
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2340/trainer_state.json
@@ -0,0 +1,2491 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.821917808219178,
+  "eval_steps": 20,
+  "global_step": 2340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.3074028554395648e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8783024645511d7f3b5e164203bd2acc49cb7792
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2360/trainer_state.json
@@ -0,0 +1,2512 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.87173100871731,
+  "eval_steps": 20,
+  "global_step": 2360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.32657339338326e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d10ba61fc725c71310384bca8d2a57f1d270af5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2380/trainer_state.json
@@ -0,0 +1,2533 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.921544209215442,
+  "eval_steps": 20,
+  "global_step": 2380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.346658566506496e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..34560b822dd26d0008a095c51faa5169115bd57c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-240/trainer_state.json
@@ -0,0 +1,286 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.597758405977584,
+  "eval_steps": 20,
+  "global_step": 240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.4041441504845824e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..85596505e1e791cd0b792a32f1a5963c9b7c0af1
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2400/trainer_state.json
@@ -0,0 +1,2554 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 5.971357409713574,
+  "eval_steps": 20,
+  "global_step": 2400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.366798492079145e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..06a2d9dcdf457f832c9d32947e477ac7d343e7a5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2420/trainer_state.json
@@ -0,0 +1,2575 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.019925280199253,
+  "eval_steps": 20,
+  "global_step": 2420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.385955790701056e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..134c2a447601c8b97d27b364cdb0568b1eb5d9ae
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2440/trainer_state.json
@@ -0,0 +1,2596 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.069738480697385,
+  "eval_steps": 20,
+  "global_step": 2440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.4063452438300467e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0340e44c48d6f6952e3c63b8d02becb8b0c886f7
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2460/trainer_state.json
@@ -0,0 +1,2617 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.119551681195516,
+  "eval_steps": 20,
+  "global_step": 2460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.4259789336409088e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf1b4f2886783238586a10fda2d4c569ffbd53aa
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2480/trainer_state.json
@@ -0,0 +1,2638 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.169364881693649,
+  "eval_steps": 20,
+  "global_step": 2480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.4449421303417446e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f03768a7c2e0d78f6be75852c13e307698fcb4a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2500/trainer_state.json
@@ -0,0 +1,2659 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.219178082191781,
+  "eval_steps": 20,
+  "global_step": 2500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.4654087754485965e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e57c3345d4fa92b2c744b636d636132d814db7df
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2520/trainer_state.json
@@ -0,0 +1,2680 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.268991282689913,
+  "eval_steps": 20,
+  "global_step": 2520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.484802811095634e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5943ab2f41c6dfff646ddf4527ebad0067111355
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2540/trainer_state.json
@@ -0,0 +1,2701 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.318804483188045,
+  "eval_steps": 20,
+  "global_step": 2540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.5037830618380902e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2dadbbdf79daa81ce42b5b584cf4d03247682c3b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2560/trainer_state.json
@@ -0,0 +1,2722 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.3686176836861765,
+  "eval_steps": 20,
+  "global_step": 2560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.5240683955550822e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8057c3d23db98004a9948a429c5dc345c3f38cba
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2580/trainer_state.json
@@ -0,0 +1,2743 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.418430884184309,
+  "eval_steps": 20,
+  "global_step": 2580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.5429140091268506e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..88ef678cb2c331a7b37839a96b8c38c3db582af8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-260/trainer_state.json
@@ -0,0 +1,307 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.6475716064757161,
+  "eval_steps": 20,
+  "global_step": 260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.606270446932787e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..794161eff33b6cee4a5de5093bbfe09c8f401050
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2600/trainer_state.json
@@ -0,0 +1,2764 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.468244084682441,
+  "eval_steps": 20,
+  "global_step": 2600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.56277747970902e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6d90725153834e1bccd36eff0820aad366b6721
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2620/trainer_state.json
@@ -0,0 +1,2785 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.518057285180573,
+  "eval_steps": 20,
+  "global_step": 2620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.5808583541512806e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..18a66dd3c51a9747d211f2886a62455b8c33f445
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2640/trainer_state.json
@@ -0,0 +1,2806 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.567870485678705,
+  "eval_steps": 20,
+  "global_step": 2640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.601385137194373e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca8c8519a5b9eab50ae333228dac51155694260e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2660/trainer_state.json
@@ -0,0 +1,2827 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.617683686176837,
+  "eval_steps": 20,
+  "global_step": 2660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.622312241424978e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..c17d9b6986952fe96bdfac1033a11db38fa6e22e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2680/trainer_state.json
@@ -0,0 +1,2848 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.667496886674969,
+  "eval_steps": 20,
+  "global_step": 2680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.6436423595865088e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..af3508a2b571bcb48eac3f53ea771b51b10964da
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2700/trainer_state.json
@@ -0,0 +1,2869 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.717310087173101,
+  "eval_steps": 20,
+  "global_step": 2700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.66178337196501e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..89e4623025da677303f3afe44ad914ae42ec18a6
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2720/trainer_state.json
@@ -0,0 +1,2890 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.767123287671232,
+  "eval_steps": 20,
+  "global_step": 2720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.679866939150684e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..400932b8647957ba324337ec57fe6f00def4a22f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2740/trainer_state.json
@@ -0,0 +1,2911 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.816936488169365,
+  "eval_steps": 20,
+  "global_step": 2740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.697999873298944e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..38604771b36d4b5160b60a5db15fe77e60a3b976
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2760/trainer_state.json
@@ -0,0 +1,2932 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.866749688667497,
+  "eval_steps": 20,
+  "global_step": 2760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.7169352449845043e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1bd7e291ee29a4d111abadb9fca4564e42b2a2cb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2780/trainer_state.json
@@ -0,0 +1,2953 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.916562889165629,
+  "eval_steps": 20,
+  "global_step": 2780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.7396395598681907e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb4955a0058e5af3ea5ce66fb0e3e13f04d1023c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-280/trainer_state.json
@@ -0,0 +1,328 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.6973848069738481,
+  "eval_steps": 20,
+  "global_step": 280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.799770988645581e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d819a80384b6cd740b55ef287795692d99d3d81c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2800/trainer_state.json
@@ -0,0 +1,2974 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6.966376089663761,
+  "eval_steps": 20,
+  "global_step": 2800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.7583451507824435e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..851edff653f4c271fc77d03dff0b9400804d4462
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2820/trainer_state.json
@@ -0,0 +1,2995 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.01494396014944,
+  "eval_steps": 20,
+  "global_step": 2820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.77769206341974e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..48fd7081598b9bc84acb0335d1db4e756c118fd2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2840/trainer_state.json
@@ -0,0 +1,3016 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.064757160647572,
+  "eval_steps": 20,
+  "global_step": 2840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.7960413146222387e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..9235afb7fc3942cd5bacfeefc1172f0681e5a974
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2860/trainer_state.json
@@ -0,0 +1,3037 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.114570361145703,
+  "eval_steps": 20,
+  "global_step": 2860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.8143420964432896e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6eab8969e6e80a494fec13597b22f2dc7af6e8a3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2880/trainer_state.json
@@ -0,0 +1,3058 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.164383561643835,
+  "eval_steps": 20,
+  "global_step": 2880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.832412199911895e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec10d0ccacb70702f638ae6fa23d1f4e06904db9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2900/trainer_state.json
@@ -0,0 +1,3079 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.214196762141968,
+  "eval_steps": 20,
+  "global_step": 2900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.8524246689629594e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..59752aa45c89b366bfe667de0a6ba217baae0e52
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2920/trainer_state.json
@@ -0,0 +1,3100 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.2640099626401,
+  "eval_steps": 20,
+  "global_step": 2920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.871313366429348e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..03a87c8182571d67db5030a88c04ee741e2abef7
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2940/trainer_state.json
@@ -0,0 +1,3121 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.313823163138232,
+  "eval_steps": 20,
+  "global_step": 2940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.893204472802079e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..147461c28c09dafcf0d6956228c24af5d8f89a22
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2960/trainer_state.json
@@ -0,0 +1,3142 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.363636363636363,
+  "eval_steps": 20,
+  "global_step": 2960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.9121676695029146e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e97c3152ee0a9a9947c7ceffc4ac08abc804972e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-2980/trainer_state.json
@@ -0,0 +1,3163 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.4134495641344955,
+  "eval_steps": 20,
+  "global_step": 2980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.930963916112097e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4b8914b46536ae633a9225167a1e49f31ad42e4
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-300/trainer_state.json
@@ -0,0 +1,349 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.7471980074719801,
+  "eval_steps": 20,
+  "global_step": 300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.9940793533825024e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e42b8003b589ecb831f900ba8fae4dd20ffad45a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3000/trainer_state.json
@@ -0,0 +1,3184 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.463262764632628,
+  "eval_steps": 20,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.949801451453624e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..827c9477245deb15154ca13769edf7d2f2f06a4e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3020/trainer_state.json
@@ -0,0 +1,3205 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.51307596513076,
+  "eval_steps": 20,
+  "global_step": 3020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.9698489261690675e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a3b7597d5e4abfe3f65724b3a535a7f95e25a22
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3040/trainer_state.json
@@ -0,0 +1,3226 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.562889165628892,
+  "eval_steps": 20,
+  "global_step": 3040,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.9910336361862554e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..aad111d71e63fd28a3c64dc820fede630a6ea9aa
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3060/trainer_state.json
@@ -0,0 +1,3247 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.6127023661270234,
+  "eval_steps": 20,
+  "global_step": 3060,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.0119894630132736e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb7d01f33862000597b365c272498ea1ae7b7fa3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3080/trainer_state.json
@@ -0,0 +1,3268 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.662515566625156,
+  "eval_steps": 20,
+  "global_step": 3080,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.0323304467608166e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..49847ab6569e8c96e7427ad867a90a238260e93e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3100/trainer_state.json
@@ -0,0 +1,3289 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.712328767123288,
+  "eval_steps": 20,
+  "global_step": 3100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.05284825399253e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..144c6b8868b24330ef04aa24aae31f60f05e04b4
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3120/trainer_state.json
@@ -0,0 +1,3310 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.76214196762142,
+  "eval_steps": 20,
+  "global_step": 3120,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.073369651548795e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c39ce1c8e08f0d4035304829c675692dcd6a86c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3140/trainer_state.json
@@ -0,0 +1,3331 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.811955168119551,
+  "eval_steps": 20,
+  "global_step": 3140,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.0946369390306714e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..87bcc17fc76a7c3a9c11e3c6724a7f20c31c112e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3160/trainer_state.json
@@ -0,0 +1,3352 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.861768368617684,
+  "eval_steps": 20,
+  "global_step": 3160,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.1162318436278886e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b89d9064479a4796df0119d2047faf327a409a7
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3180/trainer_state.json
@@ -0,0 +1,3373 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.911581569115816,
+  "eval_steps": 20,
+  "global_step": 3180,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.135096306403553e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7bca837875a8303ab6226c26c03195494af8924
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-320/trainer_state.json
@@ -0,0 +1,370 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.797011207970112,
+  "eval_steps": 20,
+  "global_step": 320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.1957299318276096e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c364a0b7a98e6300c995cc8e5240721b8d4d298
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3200/trainer_state.json
@@ -0,0 +1,3394 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.961394769613948,
+  "eval_steps": 20,
+  "global_step": 3200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.154914897928827e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f1c94b2e7e14e65ffaefacc51eed8bdeac9d2f6
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3220/trainer_state.json
@@ -0,0 +1,3415 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.009962640099626,
+  "eval_steps": 20,
+  "global_step": 3220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.174483737502474e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..761de819c6536233346326d967c160cc7303edc3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3240/trainer_state.json
@@ -0,0 +1,3436 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.059775840597759,
+  "eval_steps": 20,
+  "global_step": 3240,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.194323870975058e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c95c557a15b564259237f7ffbbd9200c59ea048
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3260/trainer_state.json
@@ -0,0 +1,3457 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.10958904109589,
+  "eval_steps": 20,
+  "global_step": 3260,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.213645202549924e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6986b2190e0607614059fa0c753f359d545ed319
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3280/trainer_state.json
@@ -0,0 +1,3478 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.159402241594023,
+  "eval_steps": 20,
+  "global_step": 3280,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.234219557393326e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..77dd688f961b0f043abfbdeed1def848a33f815b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3300/trainer_state.json
@@ -0,0 +1,3499 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.209215442092155,
+  "eval_steps": 20,
+  "global_step": 3300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.2547032565417984e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..05623ef01e4154d474e1dcf0d230602d75fdcffd
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3320/trainer_state.json
@@ -0,0 +1,3520 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.259028642590286,
+  "eval_steps": 20,
+  "global_step": 3320,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.27353450881536e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..04877e76fd7e5e9a1d3900cc9d0895ca76529ff3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3340/trainer_state.json
@@ -0,0 +1,3541 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.308841843088418,
+  "eval_steps": 20,
+  "global_step": 3340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.2929608573833626e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..af63826af1896a997e7fa614fa53b85dd79910fa
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3360/trainer_state.json
@@ -0,0 +1,3562 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.35865504358655,
+  "eval_steps": 20,
+  "global_step": 3360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.3125074818238464e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8808cce1ae7e5a9c9f59b72dcc43d8dda86a8d04
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3380/trainer_state.json
@@ -0,0 +1,3583 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.408468244084682,
+  "eval_steps": 20,
+  "global_step": 3380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.3308271128487936e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8a8f92f6eddb0fe285536241053a4a23d6163ca
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-340/trainer_state.json
@@ -0,0 +1,391 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8468244084682441,
+  "eval_steps": 20,
+  "global_step": 340,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.3829115023294464e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4112a2ec690054c3c131718081235194998057f9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3400/trainer_state.json
@@ -0,0 +1,3604 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.458281444582815,
+  "eval_steps": 20,
+  "global_step": 3400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.350161908140728e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2152e81aae17968b55755ff5a5b617ea7897765
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3420/trainer_state.json
@@ -0,0 +1,3625 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.508094645080947,
+  "eval_steps": 20,
+  "global_step": 3420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.369516450217697e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..83594e71c2103cc4968946d06831167f43be7d04
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3440/trainer_state.json
@@ -0,0 +1,3646 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.557907845579079,
+  "eval_steps": 20,
+  "global_step": 3440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.389398770003763e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..57cbd5d785876f9a12926146442ff62154971a39
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3460/trainer_state.json
@@ -0,0 +1,3667 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.607721046077211,
+  "eval_steps": 20,
+  "global_step": 3460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4093681551602074e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6dd8d013bab7f63832588a30eed9d38d9a157aa
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3480/trainer_state.json
@@ -0,0 +1,3688 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.657534246575342,
+  "eval_steps": 20,
+  "global_step": 3480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4298419809161626e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..aabab89f948543b9945b39a0f9f4d7e33f04ee12
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3500/trainer_state.json
@@ -0,0 +1,3709 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.707347447073474,
+  "eval_steps": 20,
+  "global_step": 3500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4516307630391706e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b258e6571beae878310869d2b027e68ff51f205f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3520/trainer_state.json
@@ -0,0 +1,3730 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.757160647571606,
+  "eval_steps": 20,
+  "global_step": 3520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.471473589255168e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..35e83c5ed6451f9eb6ad90860f1f6c86866a7ae4
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3540/trainer_state.json
@@ -0,0 +1,3751 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.806973848069738,
+  "eval_steps": 20,
+  "global_step": 3540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.4923127305342566e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f81c1b98940247ba821ded9d65d335eb676a6a0a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3560/trainer_state.json
@@ -0,0 +1,3772 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.85678704856787,
+  "eval_steps": 20,
+  "global_step": 3560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.509752732044042e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dcac14be3b866e8d5d0cb61348ad6bfb874260fd
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3580/trainer_state.json
@@ -0,0 +1,3793 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.906600249066003,
+  "eval_steps": 20,
+  "global_step": 3580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.527526633737134e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..fdf64e805e40024bb38a1a7e5d5ee501e4bdd63e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-360/trainer_state.json
@@ -0,0 +1,412 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.8966376089663761,
+  "eval_steps": 20,
+  "global_step": 360,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.5892743817486336e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..36ba38c374ac1c9b10cacc78177538b0919dda1d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3600/trainer_state.json
@@ -0,0 +1,3814 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 8.956413449564135,
+  "eval_steps": 20,
+  "global_step": 3600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.5487499397432525e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..e78ec1ae5aa237447b8dfdfdb7dad68f31e7b46e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3620/trainer_state.json
@@ -0,0 +1,3835 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.004981320049813,
+  "eval_steps": 20,
+  "global_step": 3620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.5680073186620416e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab57a59e2462b5fdd37d52988516edeac6a737d2
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3640/trainer_state.json
@@ -0,0 +1,3856 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.054794520547945,
+  "eval_steps": 20,
+  "global_step": 3640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.586990262147912e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c64b3a265a5f94e68edb5ba587653a5f6cd7b51
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3660/trainer_state.json
@@ -0,0 +1,3877 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.104607721046078,
+  "eval_steps": 20,
+  "global_step": 3660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.607405745129902e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc17e4c26328cb0712a3a4cdba3ce366517964f8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3680/trainer_state.json
@@ -0,0 +1,3898 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.15442092154421,
+  "eval_steps": 20,
+  "global_step": 3680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.626792600127836e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..78f950540f8a2505d5ad189e04c0ef7ffc111293
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3700/trainer_state.json
@@ -0,0 +1,3919 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.204234122042342,
+  "eval_steps": 20,
+  "global_step": 3700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.6481290013573325e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bb5fcaf47d68e20f6c6b2f88a61ffb96c50bb4c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3720/trainer_state.json
@@ -0,0 +1,3940 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.254047322540472,
+  "eval_steps": 20,
+  "global_step": 3720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.6676630596618854e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b38a24a2273be771825ae5e1b4d8b0c3cfcea025
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3740/trainer_state.json
@@ -0,0 +1,3961 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.303860523038605,
+  "eval_steps": 20,
+  "global_step": 3740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.686834495186719e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f800e5fc75df3ca79a1aa95095cb247242f2828b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3760/trainer_state.json
@@ -0,0 +1,3982 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.353673723536737,
+  "eval_steps": 20,
+  "global_step": 3760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.7059314314771046e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d336b6f7c37a357422dfe7bad46305c35a24daf
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3780/trainer_state.json
@@ -0,0 +1,4003 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.403486924034869,
+  "eval_steps": 20,
+  "global_step": 3780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.725921460999721e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..452334fbe7fa7da816637ac53ecd6c7f0aa4e466
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-380/trainer_state.json
@@ -0,0 +1,433 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9464508094645081,
+  "eval_steps": 20,
+  "global_step": 380,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.774840306202214e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..67e034590fa2e13daa2e097cd179defcff3b8a92
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3800/trainer_state.json
@@ -0,0 +1,4024 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.453300124533001,
+  "eval_steps": 20,
+  "global_step": 3800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.74812402802731e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ccb4186fbb1d1a499cb6eeb72cfab4f86f15ffec
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3820/trainer_state.json
@@ -0,0 +1,4045 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.503113325031133,
+  "eval_steps": 20,
+  "global_step": 3820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.767130308622766e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..5351f8b077de2234f95516953e85d1c747e465e0
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3840/trainer_state.json
@@ -0,0 +1,4066 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.552926525529266,
+  "eval_steps": 20,
+  "global_step": 3840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.785028076512891e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..81668befdd25e4d25e6ed4b3f3a541d4ea1fd6e7
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3860/trainer_state.json
@@ -0,0 +1,4087 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.602739726027398,
+  "eval_steps": 20,
+  "global_step": 3860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.804764090573476e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a190fdf03b113f2fe33b9dc336533aae3e620bc4
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3880/trainer_state.json
@@ -0,0 +1,4108 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.65255292652553,
+  "eval_steps": 20,
+  "global_step": 3880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    },
+    {
+      "entropy": 0.11271071215160192,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11838250607252121,
+      "learning_rate": 7.543759394953306e-07,
+      "loss": 0.05857791900634766,
+      "mean_token_accuracy": 0.9779338933527469,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.22224785959304766,
+      "eval_loss": 0.9465001225471497,
+      "eval_mean_token_accuracy": 0.8541611557783082,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.5138,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.8251104598078464e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..16383b4822b85c40feea8f95bb719ae9b112fe62
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3900/trainer_state.json
@@ -0,0 +1,4129 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.70236612702366,
+  "eval_steps": 20,
+  "global_step": 3900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    },
+    {
+      "entropy": 0.11271071215160192,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11838250607252121,
+      "learning_rate": 7.543759394953306e-07,
+      "loss": 0.05857791900634766,
+      "mean_token_accuracy": 0.9779338933527469,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.22224785959304766,
+      "eval_loss": 0.9465001225471497,
+      "eval_mean_token_accuracy": 0.8541611557783082,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.5138,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3880
+    },
+    {
+      "entropy": 0.11316167968325316,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.17645999789237976,
+      "learning_rate": 5.557293855744779e-07,
+      "loss": 0.0591269850730896,
+      "mean_token_accuracy": 0.9781935609877109,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.22210437345296838,
+      "eval_loss": 0.9471818804740906,
+      "eval_mean_token_accuracy": 0.8540260292762933,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 86.584,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.987,
+      "step": 3900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.845510683910492e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..80e3543981e81a431ac9507e73b44fdc087bbe5f
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3920/trainer_state.json
@@ -0,0 +1,4150 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.752179327521793,
+  "eval_steps": 20,
+  "global_step": 3920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    },
+    {
+      "entropy": 0.11271071215160192,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11838250607252121,
+      "learning_rate": 7.543759394953306e-07,
+      "loss": 0.05857791900634766,
+      "mean_token_accuracy": 0.9779338933527469,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.22224785959304766,
+      "eval_loss": 0.9465001225471497,
+      "eval_mean_token_accuracy": 0.8541611557783082,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.5138,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3880
+    },
+    {
+      "entropy": 0.11316167968325316,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.17645999789237976,
+      "learning_rate": 5.557293855744779e-07,
+      "loss": 0.0591269850730896,
+      "mean_token_accuracy": 0.9781935609877109,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.22210437345296838,
+      "eval_loss": 0.9471818804740906,
+      "eval_mean_token_accuracy": 0.8540260292762933,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 86.584,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.987,
+      "step": 3900
+    },
+    {
+      "entropy": 0.106583485705778,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.15395089983940125,
+      "learning_rate": 3.873080687399946e-07,
+      "loss": 0.05827186107635498,
+      "mean_token_accuracy": 0.9790284238755703,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.22207596369607505,
+      "eval_loss": 0.9475951194763184,
+      "eval_mean_token_accuracy": 0.8541645428469015,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 86.5265,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 3920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.8652538786201805e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f85138e8f2c697d61b8933db0d32daf50516b13d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3940/trainer_state.json
@@ -0,0 +1,4171 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.801992528019925,
+  "eval_steps": 20,
+  "global_step": 3940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    },
+    {
+      "entropy": 0.11271071215160192,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11838250607252121,
+      "learning_rate": 7.543759394953306e-07,
+      "loss": 0.05857791900634766,
+      "mean_token_accuracy": 0.9779338933527469,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.22224785959304766,
+      "eval_loss": 0.9465001225471497,
+      "eval_mean_token_accuracy": 0.8541611557783082,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.5138,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3880
+    },
+    {
+      "entropy": 0.11316167968325316,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.17645999789237976,
+      "learning_rate": 5.557293855744779e-07,
+      "loss": 0.0591269850730896,
+      "mean_token_accuracy": 0.9781935609877109,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.22210437345296838,
+      "eval_loss": 0.9471818804740906,
+      "eval_mean_token_accuracy": 0.8540260292762933,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 86.584,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.987,
+      "step": 3900
+    },
+    {
+      "entropy": 0.106583485705778,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.15395089983940125,
+      "learning_rate": 3.873080687399946e-07,
+      "loss": 0.05827186107635498,
+      "mean_token_accuracy": 0.9790284238755703,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.22207596369607505,
+      "eval_loss": 0.9475951194763184,
+      "eval_mean_token_accuracy": 0.8541645428469015,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 86.5265,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 3920
+    },
+    {
+      "entropy": 0.11190549619495868,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.2091812938451767,
+      "learning_rate": 2.491627825638833e-07,
+      "loss": 0.060925132036209105,
+      "mean_token_accuracy": 0.9769984439015389,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.221953320468581,
+      "eval_loss": 0.9476391077041626,
+      "eval_mean_token_accuracy": 0.854084026328353,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 86.0341,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 1.999,
+      "step": 3940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.884385820574945e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6fdfe868db1ea7d7f6b816c6fbf530863fc9bef
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3960/trainer_state.json
@@ -0,0 +1,4192 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.851805728518057,
+  "eval_steps": 20,
+  "global_step": 3960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    },
+    {
+      "entropy": 0.11271071215160192,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11838250607252121,
+      "learning_rate": 7.543759394953306e-07,
+      "loss": 0.05857791900634766,
+      "mean_token_accuracy": 0.9779338933527469,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.22224785959304766,
+      "eval_loss": 0.9465001225471497,
+      "eval_mean_token_accuracy": 0.8541611557783082,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.5138,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3880
+    },
+    {
+      "entropy": 0.11316167968325316,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.17645999789237976,
+      "learning_rate": 5.557293855744779e-07,
+      "loss": 0.0591269850730896,
+      "mean_token_accuracy": 0.9781935609877109,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.22210437345296838,
+      "eval_loss": 0.9471818804740906,
+      "eval_mean_token_accuracy": 0.8540260292762933,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 86.584,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.987,
+      "step": 3900
+    },
+    {
+      "entropy": 0.106583485705778,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.15395089983940125,
+      "learning_rate": 3.873080687399946e-07,
+      "loss": 0.05827186107635498,
+      "mean_token_accuracy": 0.9790284238755703,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.22207596369607505,
+      "eval_loss": 0.9475951194763184,
+      "eval_mean_token_accuracy": 0.8541645428469015,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 86.5265,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 3920
+    },
+    {
+      "entropy": 0.11190549619495868,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.2091812938451767,
+      "learning_rate": 2.491627825638833e-07,
+      "loss": 0.060925132036209105,
+      "mean_token_accuracy": 0.9769984439015389,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.221953320468581,
+      "eval_loss": 0.9476391077041626,
+      "eval_mean_token_accuracy": 0.854084026328353,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 86.0341,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 1.999,
+      "step": 3940
+    },
+    {
+      "entropy": 0.10218957955949008,
+      "epoch": 9.851805728518057,
+      "grad_norm": 0.12602899968624115,
+      "learning_rate": 1.4133518978026882e-07,
+      "loss": 0.053277283906936646,
+      "mean_token_accuracy": 0.9792918466031552,
+      "num_tokens": 9251912.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.851805728518057,
+      "eval_entropy": 0.22203073113463645,
+      "eval_loss": 0.9477459192276001,
+      "eval_mean_token_accuracy": 0.8541027304044989,
+      "eval_num_tokens": 9251912.0,
+      "eval_runtime": 86.6774,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.906717639286395e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..244901df63e44c7a6694193e54733a025d650fe8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-3980/trainer_state.json
@@ -0,0 +1,4213 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.90161892901619,
+  "eval_steps": 20,
+  "global_step": 3980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    },
+    {
+      "entropy": 0.11271071215160192,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11838250607252121,
+      "learning_rate": 7.543759394953306e-07,
+      "loss": 0.05857791900634766,
+      "mean_token_accuracy": 0.9779338933527469,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.22224785959304766,
+      "eval_loss": 0.9465001225471497,
+      "eval_mean_token_accuracy": 0.8541611557783082,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.5138,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3880
+    },
+    {
+      "entropy": 0.11316167968325316,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.17645999789237976,
+      "learning_rate": 5.557293855744779e-07,
+      "loss": 0.0591269850730896,
+      "mean_token_accuracy": 0.9781935609877109,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.22210437345296838,
+      "eval_loss": 0.9471818804740906,
+      "eval_mean_token_accuracy": 0.8540260292762933,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 86.584,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.987,
+      "step": 3900
+    },
+    {
+      "entropy": 0.106583485705778,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.15395089983940125,
+      "learning_rate": 3.873080687399946e-07,
+      "loss": 0.05827186107635498,
+      "mean_token_accuracy": 0.9790284238755703,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.22207596369607505,
+      "eval_loss": 0.9475951194763184,
+      "eval_mean_token_accuracy": 0.8541645428469015,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 86.5265,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 3920
+    },
+    {
+      "entropy": 0.11190549619495868,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.2091812938451767,
+      "learning_rate": 2.491627825638833e-07,
+      "loss": 0.060925132036209105,
+      "mean_token_accuracy": 0.9769984439015389,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.221953320468581,
+      "eval_loss": 0.9476391077041626,
+      "eval_mean_token_accuracy": 0.854084026328353,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 86.0341,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 1.999,
+      "step": 3940
+    },
+    {
+      "entropy": 0.10218957955949008,
+      "epoch": 9.851805728518057,
+      "grad_norm": 0.12602899968624115,
+      "learning_rate": 1.4133518978026882e-07,
+      "loss": 0.053277283906936646,
+      "mean_token_accuracy": 0.9792918466031552,
+      "num_tokens": 9251912.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.851805728518057,
+      "eval_entropy": 0.22203073113463645,
+      "eval_loss": 0.9477459192276001,
+      "eval_mean_token_accuracy": 0.8541027304044989,
+      "eval_num_tokens": 9251912.0,
+      "eval_runtime": 86.6774,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3960
+    },
+    {
+      "entropy": 0.11482724891975522,
+      "epoch": 9.90161892901619,
+      "grad_norm": 0.1705685257911682,
+      "learning_rate": 6.38578097204261e-08,
+      "loss": 0.061809581518173215,
+      "mean_token_accuracy": 0.9767013140022754,
+      "num_tokens": 9296482.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.90161892901619,
+      "eval_entropy": 0.22197876593401267,
+      "eval_loss": 0.947809636592865,
+      "eval_mean_token_accuracy": 0.8540589556444523,
+      "eval_num_tokens": 9296482.0,
+      "eval_runtime": 86.9626,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 3980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.9249780299562394e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b89a140e225d5964b2466eff7541b6f6b91c3ab9
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-40/trainer_state.json
@@ -0,0 +1,76 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.099626400996264,
+  "eval_steps": 20,
+  "global_step": 40,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3753325286326272.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2365c0d10997c131ebd642452b4239087e0d4f8d
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-400/trainer_state.json
@@ -0,0 +1,454 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.9962640099626401,
+  "eval_steps": 20,
+  "global_step": 400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.964974918647808e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae6ecebe96ebba4203052bf655333d9261a9bd80
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4000/trainer_state.json
@@ -0,0 +1,4234 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 9.951432129514322,
+  "eval_steps": 20,
+  "global_step": 4000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    },
+    {
+      "entropy": 0.11271071215160192,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11838250607252121,
+      "learning_rate": 7.543759394953306e-07,
+      "loss": 0.05857791900634766,
+      "mean_token_accuracy": 0.9779338933527469,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.22224785959304766,
+      "eval_loss": 0.9465001225471497,
+      "eval_mean_token_accuracy": 0.8541611557783082,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.5138,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3880
+    },
+    {
+      "entropy": 0.11316167968325316,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.17645999789237976,
+      "learning_rate": 5.557293855744779e-07,
+      "loss": 0.0591269850730896,
+      "mean_token_accuracy": 0.9781935609877109,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.22210437345296838,
+      "eval_loss": 0.9471818804740906,
+      "eval_mean_token_accuracy": 0.8540260292762933,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 86.584,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.987,
+      "step": 3900
+    },
+    {
+      "entropy": 0.106583485705778,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.15395089983940125,
+      "learning_rate": 3.873080687399946e-07,
+      "loss": 0.05827186107635498,
+      "mean_token_accuracy": 0.9790284238755703,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.22207596369607505,
+      "eval_loss": 0.9475951194763184,
+      "eval_mean_token_accuracy": 0.8541645428469015,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 86.5265,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 3920
+    },
+    {
+      "entropy": 0.11190549619495868,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.2091812938451767,
+      "learning_rate": 2.491627825638833e-07,
+      "loss": 0.060925132036209105,
+      "mean_token_accuracy": 0.9769984439015389,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.221953320468581,
+      "eval_loss": 0.9476391077041626,
+      "eval_mean_token_accuracy": 0.854084026328353,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 86.0341,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 1.999,
+      "step": 3940
+    },
+    {
+      "entropy": 0.10218957955949008,
+      "epoch": 9.851805728518057,
+      "grad_norm": 0.12602899968624115,
+      "learning_rate": 1.4133518978026882e-07,
+      "loss": 0.053277283906936646,
+      "mean_token_accuracy": 0.9792918466031552,
+      "num_tokens": 9251912.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.851805728518057,
+      "eval_entropy": 0.22203073113463645,
+      "eval_loss": 0.9477459192276001,
+      "eval_mean_token_accuracy": 0.8541027304044989,
+      "eval_num_tokens": 9251912.0,
+      "eval_runtime": 86.6774,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3960
+    },
+    {
+      "entropy": 0.11482724891975522,
+      "epoch": 9.90161892901619,
+      "grad_norm": 0.1705685257911682,
+      "learning_rate": 6.38578097204261e-08,
+      "loss": 0.061809581518173215,
+      "mean_token_accuracy": 0.9767013140022754,
+      "num_tokens": 9296482.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.90161892901619,
+      "eval_entropy": 0.22197876593401267,
+      "eval_loss": 0.947809636592865,
+      "eval_mean_token_accuracy": 0.8540589556444523,
+      "eval_num_tokens": 9296482.0,
+      "eval_runtime": 86.9626,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 3980
+    },
+    {
+      "entropy": 0.1206501518841833,
+      "epoch": 9.951432129514322,
+      "grad_norm": 0.17280042171478271,
+      "learning_rate": 1.675400850544711e-08,
+      "loss": 0.06324169635772706,
+      "mean_token_accuracy": 0.9753918401896954,
+      "num_tokens": 9341118.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.951432129514322,
+      "eval_entropy": 0.22201869714745257,
+      "eval_loss": 0.9477588534355164,
+      "eval_mean_token_accuracy": 0.8541659501402877,
+      "eval_num_tokens": 9341118.0,
+      "eval_runtime": 86.6633,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 4000
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.9449716498034074e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..43b44e849978a49365c74c5621bf6b8058795330
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-4020/trainer_state.json
@@ -0,0 +1,4255 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 10.0,
+  "eval_steps": 20,
+  "global_step": 4020,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    },
+    {
+      "entropy": 0.4727417893707752,
+      "epoch": 2.488169364881694,
+      "grad_norm": 0.7852500677108765,
+      "learning_rate": 0.0001883129646126818,
+      "loss": 0.4142886161804199,
+      "mean_token_accuracy": 0.8712429471313954,
+      "num_tokens": 2333733.0,
+      "step": 1000
+    },
+    {
+      "epoch": 2.488169364881694,
+      "eval_entropy": 0.5386548059624295,
+      "eval_loss": 0.536101222038269,
+      "eval_mean_token_accuracy": 0.8499491239009902,
+      "eval_num_tokens": 2333733.0,
+      "eval_runtime": 86.9501,
+      "eval_samples_per_second": 15.825,
+      "eval_steps_per_second": 1.978,
+      "step": 1000
+    },
+    {
+      "entropy": 0.4673406321555376,
+      "epoch": 2.5379825653798256,
+      "grad_norm": 0.7133921384811401,
+      "learning_rate": 0.0001874326505341286,
+      "loss": 0.40857529640197754,
+      "mean_token_accuracy": 0.8747925907373428,
+      "num_tokens": 2384270.0,
+      "step": 1020
+    },
+    {
+      "epoch": 2.5379825653798256,
+      "eval_entropy": 0.495788364909416,
+      "eval_loss": 0.5418923497200012,
+      "eval_mean_token_accuracy": 0.851321972040243,
+      "eval_num_tokens": 2384270.0,
+      "eval_runtime": 86.7154,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.983,
+      "step": 1020
+    },
+    {
+      "entropy": 0.47599745728075504,
+      "epoch": 2.587795765877958,
+      "grad_norm": 0.8202953338623047,
+      "learning_rate": 0.0001865262021621137,
+      "loss": 0.40998234748840334,
+      "mean_token_accuracy": 0.8758242674171924,
+      "num_tokens": 2428036.0,
+      "step": 1040
+    },
+    {
+      "epoch": 2.587795765877958,
+      "eval_entropy": 0.4887966953737791,
+      "eval_loss": 0.5408804416656494,
+      "eval_mean_token_accuracy": 0.8512661065473113,
+      "eval_num_tokens": 2428036.0,
+      "eval_runtime": 86.7869,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 1040
+    },
+    {
+      "entropy": 0.4824396539479494,
+      "epoch": 2.6376089663760895,
+      "grad_norm": 0.6507360935211182,
+      "learning_rate": 0.00018559389286910275,
+      "loss": 0.4165764808654785,
+      "mean_token_accuracy": 0.8722914069890976,
+      "num_tokens": 2476815.0,
+      "step": 1060
+    },
+    {
+      "epoch": 2.6376089663760895,
+      "eval_entropy": 0.4793398808254752,
+      "eval_loss": 0.5326959490776062,
+      "eval_mean_token_accuracy": 0.8534493650807891,
+      "eval_num_tokens": 2476815.0,
+      "eval_runtime": 86.9559,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1060
+    },
+    {
+      "entropy": 0.4605010639876127,
+      "epoch": 2.6874221668742218,
+      "grad_norm": 0.6740535497665405,
+      "learning_rate": 0.00018463600382686253,
+      "loss": 0.4123940944671631,
+      "mean_token_accuracy": 0.8733638986945153,
+      "num_tokens": 2527131.0,
+      "step": 1080
+    },
+    {
+      "epoch": 2.6874221668742218,
+      "eval_entropy": 0.47902208583992584,
+      "eval_loss": 0.5372340083122253,
+      "eval_mean_token_accuracy": 0.851325950303743,
+      "eval_num_tokens": 2527131.0,
+      "eval_runtime": 86.9638,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 1080
+    },
+    {
+      "entropy": 0.4872019402682781,
+      "epoch": 2.7372353673723535,
+      "grad_norm": 0.6994742155075073,
+      "learning_rate": 0.0001836528239216632,
+      "loss": 0.41599602699279786,
+      "mean_token_accuracy": 0.872775862365961,
+      "num_tokens": 2572537.0,
+      "step": 1100
+    },
+    {
+      "epoch": 2.7372353673723535,
+      "eval_entropy": 0.4893243626453156,
+      "eval_loss": 0.5327795743942261,
+      "eval_mean_token_accuracy": 0.8537560302850812,
+      "eval_num_tokens": 2572537.0,
+      "eval_runtime": 86.823,
+      "eval_samples_per_second": 15.848,
+      "eval_steps_per_second": 1.981,
+      "step": 1100
+    },
+    {
+      "entropy": 0.4949610233306885,
+      "epoch": 2.7870485678704857,
+      "grad_norm": 0.9605912566184998,
+      "learning_rate": 0.0001826446496671543,
+      "loss": 0.4266993045806885,
+      "mean_token_accuracy": 0.8688005246222019,
+      "num_tokens": 2616047.0,
+      "step": 1120
+    },
+    {
+      "epoch": 2.7870485678704857,
+      "eval_entropy": 0.5061517927882283,
+      "eval_loss": 0.5356810092926025,
+      "eval_mean_token_accuracy": 0.8524593568818514,
+      "eval_num_tokens": 2616047.0,
+      "eval_runtime": 86.8385,
+      "eval_samples_per_second": 15.846,
+      "eval_steps_per_second": 1.981,
+      "step": 1120
+    },
+    {
+      "entropy": 0.48401356525719164,
+      "epoch": 2.8368617683686175,
+      "grad_norm": 0.6332499980926514,
+      "learning_rate": 0.00018161178511494022,
+      "loss": 0.42131738662719725,
+      "mean_token_accuracy": 0.8729447312653065,
+      "num_tokens": 2664744.0,
+      "step": 1140
+    },
+    {
+      "epoch": 2.8368617683686175,
+      "eval_entropy": 0.48792897060860035,
+      "eval_loss": 0.5290402173995972,
+      "eval_mean_token_accuracy": 0.853078076659247,
+      "eval_num_tokens": 2664744.0,
+      "eval_runtime": 86.7746,
+      "eval_samples_per_second": 15.857,
+      "eval_steps_per_second": 1.982,
+      "step": 1140
+    },
+    {
+      "entropy": 0.48017631396651267,
+      "epoch": 2.8866749688667497,
+      "grad_norm": 0.8013222217559814,
+      "learning_rate": 0.00018055454176288234,
+      "loss": 0.4195223808288574,
+      "mean_token_accuracy": 0.8721428856253624,
+      "num_tokens": 2710196.0,
+      "step": 1160
+    },
+    {
+      "epoch": 2.8866749688667497,
+      "eval_entropy": 0.5205728571082271,
+      "eval_loss": 0.5242091417312622,
+      "eval_mean_token_accuracy": 0.8535184077052183,
+      "eval_num_tokens": 2710196.0,
+      "eval_runtime": 87.009,
+      "eval_samples_per_second": 15.814,
+      "eval_steps_per_second": 1.977,
+      "step": 1160
+    },
+    {
+      "entropy": 0.4914963707327843,
+      "epoch": 2.936488169364882,
+      "grad_norm": 0.7638444900512695,
+      "learning_rate": 0.00017947323846115553,
+      "loss": 0.42672080993652345,
+      "mean_token_accuracy": 0.8699039086699486,
+      "num_tokens": 2758752.0,
+      "step": 1180
+    },
+    {
+      "epoch": 2.936488169364882,
+      "eval_entropy": 0.5012746392640957,
+      "eval_loss": 0.523289144039154,
+      "eval_mean_token_accuracy": 0.8543414289868155,
+      "eval_num_tokens": 2758752.0,
+      "eval_runtime": 86.939,
+      "eval_samples_per_second": 15.827,
+      "eval_steps_per_second": 1.978,
+      "step": 1180
+    },
+    {
+      "entropy": 0.48674912229180334,
+      "epoch": 2.9863013698630136,
+      "grad_norm": 0.6185152530670166,
+      "learning_rate": 0.00017836820131608745,
+      "loss": 0.4154937744140625,
+      "mean_token_accuracy": 0.8686325438320637,
+      "num_tokens": 2803186.0,
+      "step": 1200
+    },
+    {
+      "epoch": 2.9863013698630136,
+      "eval_entropy": 0.4763167448515116,
+      "eval_loss": 0.5311836004257202,
+      "eval_mean_token_accuracy": 0.8521223698937616,
+      "eval_num_tokens": 2803186.0,
+      "eval_runtime": 86.6775,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 1200
+    },
+    {
+      "entropy": 0.40521582922874355,
+      "epoch": 3.0348692403486925,
+      "grad_norm": 0.7635012269020081,
+      "learning_rate": 0.0001772397635918091,
+      "loss": 0.3426941394805908,
+      "mean_token_accuracy": 0.8914182109710498,
+      "num_tokens": 2848509.0,
+      "step": 1220
+    },
+    {
+      "epoch": 3.0348692403486925,
+      "eval_entropy": 0.444144579733527,
+      "eval_loss": 0.5566298961639404,
+      "eval_mean_token_accuracy": 0.8528384368780048,
+      "eval_num_tokens": 2848509.0,
+      "eval_runtime": 87.0782,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 1220
+    },
+    {
+      "entropy": 0.3624659780412912,
+      "epoch": 3.0846824408468243,
+      "grad_norm": 0.7089799046516418,
+      "learning_rate": 0.000176088265609747,
+      "loss": 0.29830265045166016,
+      "mean_token_accuracy": 0.9024251900613308,
+      "num_tokens": 2894976.0,
+      "step": 1240
+    },
+    {
+      "epoch": 3.0846824408468243,
+      "eval_entropy": 0.4600461038739182,
+      "eval_loss": 0.5547332167625427,
+      "eval_mean_token_accuracy": 0.8523849744436353,
+      "eval_num_tokens": 2894976.0,
+      "eval_runtime": 86.8998,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1240
+    },
+    {
+      "entropy": 0.3730695676058531,
+      "epoch": 3.1344956413449565,
+      "grad_norm": 0.850790798664093,
+      "learning_rate": 0.00017491405464598693,
+      "loss": 0.30087215900421144,
+      "mean_token_accuracy": 0.9013829864561558,
+      "num_tokens": 2944401.0,
+      "step": 1260
+    },
+    {
+      "epoch": 3.1344956413449565,
+      "eval_entropy": 0.43883725758208786,
+      "eval_loss": 0.556238055229187,
+      "eval_mean_token_accuracy": 0.851823707999185,
+      "eval_num_tokens": 2944401.0,
+      "eval_runtime": 86.6609,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1260
+    },
+    {
+      "entropy": 0.3766512680798769,
+      "epoch": 3.1843088418430883,
+      "grad_norm": 0.8249748945236206,
+      "learning_rate": 0.00017371748482654015,
+      "loss": 0.3063816070556641,
+      "mean_token_accuracy": 0.9013078331947326,
+      "num_tokens": 2990243.0,
+      "step": 1280
+    },
+    {
+      "epoch": 3.1843088418430883,
+      "eval_entropy": 0.455615431763405,
+      "eval_loss": 0.5655665993690491,
+      "eval_mean_token_accuracy": 0.8502279263596202,
+      "eval_num_tokens": 2990243.0,
+      "eval_runtime": 86.9248,
+      "eval_samples_per_second": 15.83,
+      "eval_steps_per_second": 1.979,
+      "step": 1280
+    },
+    {
+      "entropy": 0.38341562673449514,
+      "epoch": 3.2341220423412205,
+      "grad_norm": 0.7116150856018066,
+      "learning_rate": 0.0001724989170205438,
+      "loss": 0.3090095043182373,
+      "mean_token_accuracy": 0.8975317768752575,
+      "num_tokens": 3037248.0,
+      "step": 1300
+    },
+    {
+      "epoch": 3.2341220423412205,
+      "eval_entropy": 0.43765638578076693,
+      "eval_loss": 0.5622899532318115,
+      "eval_mean_token_accuracy": 0.8535801556914352,
+      "eval_num_tokens": 3037248.0,
+      "eval_runtime": 86.8191,
+      "eval_samples_per_second": 15.849,
+      "eval_steps_per_second": 1.981,
+      "step": 1300
+    },
+    {
+      "entropy": 0.3648845721036196,
+      "epoch": 3.2839352428393527,
+      "grad_norm": 2.4296491146087646,
+      "learning_rate": 0.0001712587187314278,
+      "loss": 0.30108642578125,
+      "mean_token_accuracy": 0.9024063929915428,
+      "num_tokens": 3088676.0,
+      "step": 1320
+    },
+    {
+      "epoch": 3.2839352428393527,
+      "eval_entropy": 0.47450404683517855,
+      "eval_loss": 0.5497040152549744,
+      "eval_mean_token_accuracy": 0.852605615590894,
+      "eval_num_tokens": 3088676.0,
+      "eval_runtime": 86.8508,
+      "eval_samples_per_second": 15.843,
+      "eval_steps_per_second": 1.98,
+      "step": 1320
+    },
+    {
+      "entropy": 0.40571765452623365,
+      "epoch": 3.3337484433374844,
+      "grad_norm": 0.8728042244911194,
+      "learning_rate": 0.00016999726398608065,
+      "loss": 0.3268932104110718,
+      "mean_token_accuracy": 0.8947565704584122,
+      "num_tokens": 3132874.0,
+      "step": 1340
+    },
+    {
+      "epoch": 3.3337484433374844,
+      "eval_entropy": 0.45309220496998276,
+      "eval_loss": 0.5505508780479431,
+      "eval_mean_token_accuracy": 0.8525978032239648,
+      "eval_num_tokens": 3132874.0,
+      "eval_runtime": 87.0437,
+      "eval_samples_per_second": 15.808,
+      "eval_steps_per_second": 1.976,
+      "step": 1340
+    },
+    {
+      "entropy": 0.39150902144610883,
+      "epoch": 3.383561643835616,
+      "grad_norm": 0.9764124155044556,
+      "learning_rate": 0.00016871493322204826,
+      "loss": 0.32332139015197753,
+      "mean_token_accuracy": 0.8968119442462921,
+      "num_tokens": 3178055.0,
+      "step": 1360
+    },
+    {
+      "epoch": 3.383561643835616,
+      "eval_entropy": 0.4515272902194844,
+      "eval_loss": 0.5608078837394714,
+      "eval_mean_token_accuracy": 0.8525907397270203,
+      "eval_num_tokens": 3178055.0,
+      "eval_runtime": 87.1423,
+      "eval_samples_per_second": 15.79,
+      "eval_steps_per_second": 1.974,
+      "step": 1360
+    },
+    {
+      "entropy": 0.39573758877813814,
+      "epoch": 3.4333748443337484,
+      "grad_norm": 0.7583338618278503,
+      "learning_rate": 0.00016741211317279893,
+      "loss": 0.3187565803527832,
+      "mean_token_accuracy": 0.8971615083515644,
+      "num_tokens": 3223389.0,
+      "step": 1380
+    },
+    {
+      "epoch": 3.4333748443337484,
+      "eval_entropy": 0.4382614852037541,
+      "eval_loss": 0.5680410861968994,
+      "eval_mean_token_accuracy": 0.851775293433389,
+      "eval_num_tokens": 3223389.0,
+      "eval_runtime": 86.9039,
+      "eval_samples_per_second": 15.834,
+      "eval_steps_per_second": 1.979,
+      "step": 1380
+    },
+    {
+      "entropy": 0.38341186530888083,
+      "epoch": 3.4831880448318806,
+      "grad_norm": 0.8511857986450195,
+      "learning_rate": 0.00016608919675108996,
+      "loss": 0.31606297492980956,
+      "mean_token_accuracy": 0.8974148526787757,
+      "num_tokens": 3274096.0,
+      "step": 1400
+    },
+    {
+      "epoch": 3.4831880448318806,
+      "eval_entropy": 0.42544752082159354,
+      "eval_loss": 0.5536707639694214,
+      "eval_mean_token_accuracy": 0.854495684074801,
+      "eval_num_tokens": 3274096.0,
+      "eval_runtime": 87.0017,
+      "eval_samples_per_second": 15.816,
+      "eval_steps_per_second": 1.977,
+      "step": 1400
+    },
+    {
+      "entropy": 0.3771478392183781,
+      "epoch": 3.5330012453300124,
+      "grad_norm": 0.8417750000953674,
+      "learning_rate": 0.0001647465829304705,
+      "loss": 0.3137716054916382,
+      "mean_token_accuracy": 0.900364401191473,
+      "num_tokens": 3322044.0,
+      "step": 1420
+    },
+    {
+      "epoch": 3.5330012453300124,
+      "eval_entropy": 0.4379603945931723,
+      "eval_loss": 0.5532916784286499,
+      "eval_mean_token_accuracy": 0.8527980658204056,
+      "eval_num_tokens": 3322044.0,
+      "eval_runtime": 86.9545,
+      "eval_samples_per_second": 15.824,
+      "eval_steps_per_second": 1.978,
+      "step": 1420
+    },
+    {
+      "entropy": 0.39049469511955975,
+      "epoch": 3.5828144458281446,
+      "grad_norm": 0.7178560495376587,
+      "learning_rate": 0.0001633846766249565,
+      "loss": 0.32054433822631834,
+      "mean_token_accuracy": 0.8984041027724743,
+      "num_tokens": 3369147.0,
+      "step": 1440
+    },
+    {
+      "epoch": 3.5828144458281446,
+      "eval_entropy": 0.44776528259349424,
+      "eval_loss": 0.5596351027488708,
+      "eval_mean_token_accuracy": 0.8523151042849518,
+      "eval_num_tokens": 3369147.0,
+      "eval_runtime": 86.6601,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 1440
+    },
+    {
+      "entropy": 0.3994994066655636,
+      "epoch": 3.6326276463262763,
+      "grad_norm": 0.6914435625076294,
+      "learning_rate": 0.00016200388856691433,
+      "loss": 0.32937595844268797,
+      "mean_token_accuracy": 0.893961051106453,
+      "num_tokens": 3415600.0,
+      "step": 1460
+    },
+    {
+      "epoch": 3.6326276463262763,
+      "eval_entropy": 0.4416292599120805,
+      "eval_loss": 0.5500746965408325,
+      "eval_mean_token_accuracy": 0.852156569444856,
+      "eval_num_tokens": 3415600.0,
+      "eval_runtime": 87.1683,
+      "eval_samples_per_second": 15.786,
+      "eval_steps_per_second": 1.973,
+      "step": 1460
+    },
+    {
+      "entropy": 0.40447953082621096,
+      "epoch": 3.6824408468244085,
+      "grad_norm": 0.669640839099884,
+      "learning_rate": 0.00016060463518318926,
+      "loss": 0.32666375637054446,
+      "mean_token_accuracy": 0.8955025888979435,
+      "num_tokens": 3460471.0,
+      "step": 1480
+    },
+    {
+      "epoch": 3.6824408468244085,
+      "eval_entropy": 0.44157897559709325,
+      "eval_loss": 0.555138111114502,
+      "eval_mean_token_accuracy": 0.8522906923709914,
+      "eval_num_tokens": 3460471.0,
+      "eval_runtime": 87.3983,
+      "eval_samples_per_second": 15.744,
+      "eval_steps_per_second": 1.968,
+      "step": 1480
+    },
+    {
+      "entropy": 0.3910129055380821,
+      "epoch": 3.7322540473225407,
+      "grad_norm": 0.6813255548477173,
+      "learning_rate": 0.00015918733846951693,
+      "loss": 0.3226655960083008,
+      "mean_token_accuracy": 0.8959677331149578,
+      "num_tokens": 3507647.0,
+      "step": 1500
+    },
+    {
+      "epoch": 3.7322540473225407,
+      "eval_entropy": 0.4205027032037114,
+      "eval_loss": 0.5615105628967285,
+      "eval_mean_token_accuracy": 0.852928082263747,
+      "eval_num_tokens": 3507647.0,
+      "eval_runtime": 87.1162,
+      "eval_samples_per_second": 15.795,
+      "eval_steps_per_second": 1.974,
+      "step": 1500
+    },
+    {
+      "entropy": 0.3978020466864109,
+      "epoch": 3.7820672478206725,
+      "grad_norm": 0.7461448907852173,
+      "learning_rate": 0.00015775242586325538,
+      "loss": 0.33161282539367676,
+      "mean_token_accuracy": 0.8929055534303189,
+      "num_tokens": 3548934.0,
+      "step": 1520
+    },
+    {
+      "epoch": 3.7820672478206725,
+      "eval_entropy": 0.42685241338818575,
+      "eval_loss": 0.558264970779419,
+      "eval_mean_token_accuracy": 0.8518715078054473,
+      "eval_num_tokens": 3548934.0,
+      "eval_runtime": 87.1858,
+      "eval_samples_per_second": 15.782,
+      "eval_steps_per_second": 1.973,
+      "step": 1520
+    },
+    {
+      "entropy": 0.38580134473741057,
+      "epoch": 3.8318804483188043,
+      "grad_norm": 0.8667106032371521,
+      "learning_rate": 0.00015630033011447534,
+      "loss": 0.3317128896713257,
+      "mean_token_accuracy": 0.8927417695522308,
+      "num_tokens": 3597186.0,
+      "step": 1540
+    },
+    {
+      "epoch": 3.8318804483188043,
+      "eval_entropy": 0.44643741187661196,
+      "eval_loss": 0.5528179407119751,
+      "eval_mean_token_accuracy": 0.8526672790216845,
+      "eval_num_tokens": 3597186.0,
+      "eval_runtime": 87.1379,
+      "eval_samples_per_second": 15.791,
+      "eval_steps_per_second": 1.974,
+      "step": 1540
+    },
+    {
+      "entropy": 0.38523379489779475,
+      "epoch": 3.8816936488169365,
+      "grad_norm": 1.001920461654663,
+      "learning_rate": 0.0001548314891554491,
+      "loss": 0.3263416051864624,
+      "mean_token_accuracy": 0.8965372897684574,
+      "num_tokens": 3646646.0,
+      "step": 1560
+    },
+    {
+      "epoch": 3.8816936488169365,
+      "eval_entropy": 0.4217506472107976,
+      "eval_loss": 0.5582898855209351,
+      "eval_mean_token_accuracy": 0.8525883361350658,
+      "eval_num_tokens": 3646646.0,
+      "eval_runtime": 87.177,
+      "eval_samples_per_second": 15.784,
+      "eval_steps_per_second": 1.973,
+      "step": 1560
+    },
+    {
+      "entropy": 0.39575362466275693,
+      "epoch": 3.9315068493150687,
+      "grad_norm": 0.7755498290061951,
+      "learning_rate": 0.00015334634596857598,
+      "loss": 0.32946195602416994,
+      "mean_token_accuracy": 0.8922067232429981,
+      "num_tokens": 3689560.0,
+      "step": 1580
+    },
+    {
+      "epoch": 3.9315068493150687,
+      "eval_entropy": 0.44022568661806194,
+      "eval_loss": 0.5536255240440369,
+      "eval_mean_token_accuracy": 0.853287082078845,
+      "eval_num_tokens": 3689560.0,
+      "eval_runtime": 87.0747,
+      "eval_samples_per_second": 15.803,
+      "eval_steps_per_second": 1.975,
+      "step": 1580
+    },
+    {
+      "entropy": 0.38972948826849463,
+      "epoch": 3.9813200498132004,
+      "grad_norm": 0.9432196617126465,
+      "learning_rate": 0.0001518453484527848,
+      "loss": 0.3292489767074585,
+      "mean_token_accuracy": 0.8934461928904056,
+      "num_tokens": 3736533.0,
+      "step": 1600
+    },
+    {
+      "epoch": 3.9813200498132004,
+      "eval_entropy": 0.43847283859585606,
+      "eval_loss": 0.5473363995552063,
+      "eval_mean_token_accuracy": 0.853469139268232,
+      "eval_num_tokens": 3736533.0,
+      "eval_runtime": 86.8446,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.981,
+      "step": 1600
+    },
+    {
+      "entropy": 0.3139461097426904,
+      "epoch": 4.029887920298879,
+      "grad_norm": 0.9580853581428528,
+      "learning_rate": 0.00015032894928845378,
+      "loss": 0.2577517986297607,
+      "mean_token_accuracy": 0.9184213326527522,
+      "num_tokens": 3783985.0,
+      "step": 1620
+    },
+    {
+      "epoch": 4.029887920298879,
+      "eval_entropy": 0.36030947936828744,
+      "eval_loss": 0.6320357322692871,
+      "eval_mean_token_accuracy": 0.8514543522235959,
+      "eval_num_tokens": 3783985.0,
+      "eval_runtime": 86.4872,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1620
+    },
+    {
+      "entropy": 0.283712088316679,
+      "epoch": 4.0797011207970115,
+      "grad_norm": 0.6759635210037231,
+      "learning_rate": 0.00014879760580088826,
+      "loss": 0.21372199058532715,
+      "mean_token_accuracy": 0.9291425101459027,
+      "num_tokens": 3828830.0,
+      "step": 1640
+    },
+    {
+      "epoch": 4.0797011207970115,
+      "eval_entropy": 0.3795742853436359,
+      "eval_loss": 0.6112104654312134,
+      "eval_mean_token_accuracy": 0.8501740534638249,
+      "eval_num_tokens": 3828830.0,
+      "eval_runtime": 86.7132,
+      "eval_samples_per_second": 15.868,
+      "eval_steps_per_second": 1.984,
+      "step": 1640
+    },
+    {
+      "entropy": 0.2836189458146691,
+      "epoch": 4.129514321295143,
+      "grad_norm": 0.6797131299972534,
+      "learning_rate": 0.00014725177982239767,
+      "loss": 0.21063549518585206,
+      "mean_token_accuracy": 0.9303271003067494,
+      "num_tokens": 3878616.0,
+      "step": 1660
+    },
+    {
+      "epoch": 4.129514321295143,
+      "eval_entropy": 0.38122218240832173,
+      "eval_loss": 0.6002676486968994,
+      "eval_mean_token_accuracy": 0.8516882522854694,
+      "eval_num_tokens": 3878616.0,
+      "eval_runtime": 86.7816,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 1660
+    },
+    {
+      "entropy": 0.2938332514837384,
+      "epoch": 4.179327521793275,
+      "grad_norm": 2.0500826835632324,
+      "learning_rate": 0.00014569193755301305,
+      "loss": 0.2256624460220337,
+      "mean_token_accuracy": 0.9247965328395367,
+      "num_tokens": 3924763.0,
+      "step": 1680
+    },
+    {
+      "epoch": 4.179327521793275,
+      "eval_entropy": 0.38222643785005395,
+      "eval_loss": 0.6075488924980164,
+      "eval_mean_token_accuracy": 0.8500458916952444,
+      "eval_num_tokens": 3924763.0,
+      "eval_runtime": 86.7462,
+      "eval_samples_per_second": 15.862,
+      "eval_steps_per_second": 1.983,
+      "step": 1680
+    },
+    {
+      "entropy": 0.2907615687698126,
+      "epoch": 4.229140722291407,
+      "grad_norm": 0.6996814608573914,
+      "learning_rate": 0.00014411854941988772,
+      "loss": 0.2183893918991089,
+      "mean_token_accuracy": 0.9268265649676323,
+      "num_tokens": 3967474.0,
+      "step": 1700
+    },
+    {
+      "epoch": 4.229140722291407,
+      "eval_entropy": 0.3791500868838887,
+      "eval_loss": 0.6009940505027771,
+      "eval_mean_token_accuracy": 0.8522552584492883,
+      "eval_num_tokens": 3967474.0,
+      "eval_runtime": 86.6182,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 1700
+    },
+    {
+      "entropy": 0.2931097410619259,
+      "epoch": 4.2789539227895395,
+      "grad_norm": 1.047773003578186,
+      "learning_rate": 0.00014253208993542275,
+      "loss": 0.22391772270202637,
+      "mean_token_accuracy": 0.9256755143404007,
+      "num_tokens": 4013170.0,
+      "step": 1720
+    },
+    {
+      "epoch": 4.2789539227895395,
+      "eval_entropy": 0.3881096559208493,
+      "eval_loss": 0.6058024764060974,
+      "eval_mean_token_accuracy": 0.8516207212625548,
+      "eval_num_tokens": 4013170.0,
+      "eval_runtime": 86.7613,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.982,
+      "step": 1720
+    },
+    {
+      "entropy": 0.2862068932503462,
+      "epoch": 4.328767123287671,
+      "grad_norm": 0.7637642025947571,
+      "learning_rate": 0.00014093303755416048,
+      "loss": 0.21297969818115234,
+      "mean_token_accuracy": 0.9277966007590294,
+      "num_tokens": 4062594.0,
+      "step": 1740
+    },
+    {
+      "epoch": 4.328767123287671,
+      "eval_entropy": 0.3942466052465661,
+      "eval_loss": 0.596074104309082,
+      "eval_mean_token_accuracy": 0.8516605569179668,
+      "eval_num_tokens": 4062594.0,
+      "eval_runtime": 86.4379,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 1740
+    },
+    {
+      "entropy": 0.28564709294587376,
+      "epoch": 4.378580323785803,
+      "grad_norm": 0.9261002540588379,
+      "learning_rate": 0.00013932187452848907,
+      "loss": 0.21695659160614014,
+      "mean_token_accuracy": 0.9291524566709995,
+      "num_tokens": 4110719.0,
+      "step": 1760
+    },
+    {
+      "epoch": 4.378580323785803,
+      "eval_entropy": 0.38845692921516506,
+      "eval_loss": 0.6026391983032227,
+      "eval_mean_token_accuracy": 0.8520834189514781,
+      "eval_num_tokens": 4110719.0,
+      "eval_runtime": 86.6486,
+      "eval_samples_per_second": 15.88,
+      "eval_steps_per_second": 1.985,
+      "step": 1760
+    },
+    {
+      "entropy": 0.29184935465455053,
+      "epoch": 4.428393524283935,
+      "grad_norm": 0.8760656714439392,
+      "learning_rate": 0.00013769908676320198,
+      "loss": 0.21900837421417235,
+      "mean_token_accuracy": 0.9270037837326527,
+      "num_tokens": 4156827.0,
+      "step": 1780
+    },
+    {
+      "epoch": 4.428393524283935,
+      "eval_entropy": 0.372284933404867,
+      "eval_loss": 0.6095370054244995,
+      "eval_mean_token_accuracy": 0.8511590410110562,
+      "eval_num_tokens": 4156827.0,
+      "eval_runtime": 86.6256,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 1780
+    },
+    {
+      "entropy": 0.30764632020145655,
+      "epoch": 4.478206724782067,
+      "grad_norm": 0.7438597679138184,
+      "learning_rate": 0.00013606516366895552,
+      "loss": 0.23197441101074218,
+      "mean_token_accuracy": 0.922636279463768,
+      "num_tokens": 4200878.0,
+      "step": 1800
+    },
+    {
+      "epoch": 4.478206724782067,
+      "eval_entropy": 0.37462779758281484,
+      "eval_loss": 0.6045190691947937,
+      "eval_mean_token_accuracy": 0.851631471237471,
+      "eval_num_tokens": 4200878.0,
+      "eval_runtime": 86.457,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 1800
+    },
+    {
+      "entropy": 0.3059290213510394,
+      "epoch": 4.5280199252802,
+      "grad_norm": 0.7649065852165222,
+      "learning_rate": 0.00013442059801466934,
+      "loss": 0.22677943706512452,
+      "mean_token_accuracy": 0.9249936550855636,
+      "num_tokens": 4244423.0,
+      "step": 1820
+    },
+    {
+      "epoch": 4.5280199252802,
+      "eval_entropy": 0.38022891068181325,
+      "eval_loss": 0.5899756550788879,
+      "eval_mean_token_accuracy": 0.8530971352444139,
+      "eval_num_tokens": 4244423.0,
+      "eval_runtime": 86.7306,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 1820
+    },
+    {
+      "entropy": 0.30684010963886976,
+      "epoch": 4.577833125778331,
+      "grad_norm": 0.6305304169654846,
+      "learning_rate": 0.00013276588577891408,
+      "loss": 0.2301240921020508,
+      "mean_token_accuracy": 0.9220072351396084,
+      "num_tokens": 4289773.0,
+      "step": 1840
+    },
+    {
+      "epoch": 4.577833125778331,
+      "eval_entropy": 0.3938776440052099,
+      "eval_loss": 0.6018155813217163,
+      "eval_mean_token_accuracy": 0.8511968948120294,
+      "eval_num_tokens": 4289773.0,
+      "eval_runtime": 86.5261,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 1840
+    },
+    {
+      "entropy": 0.29356779605150224,
+      "epoch": 4.627646326276463,
+      "grad_norm": 0.8183266520500183,
+      "learning_rate": 0.00013110152600033086,
+      "loss": 0.2146066188812256,
+      "mean_token_accuracy": 0.9264136716723442,
+      "num_tokens": 4336894.0,
+      "step": 1860
+    },
+    {
+      "epoch": 4.627646326276463,
+      "eval_entropy": 0.3670051993672238,
+      "eval_loss": 0.6133230924606323,
+      "eval_mean_token_accuracy": 0.8523070160732713,
+      "eval_num_tokens": 4336894.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 1860
+    },
+    {
+      "entropy": 0.30114926025271416,
+      "epoch": 4.677459526774595,
+      "grad_norm": 0.7963266372680664,
+      "learning_rate": 0.00012942802062712827,
+      "loss": 0.2227048397064209,
+      "mean_token_accuracy": 0.9253233529627323,
+      "num_tokens": 4382947.0,
+      "step": 1880
+    },
+    {
+      "epoch": 4.677459526774595,
+      "eval_entropy": 0.3668888188725294,
+      "eval_loss": 0.6080331802368164,
+      "eval_mean_token_accuracy": 0.8522803031427916,
+      "eval_num_tokens": 4382947.0,
+      "eval_runtime": 86.3427,
+      "eval_samples_per_second": 15.937,
+      "eval_steps_per_second": 1.992,
+      "step": 1880
+    },
+    {
+      "entropy": 0.28620288632810115,
+      "epoch": 4.7272727272727275,
+      "grad_norm": 0.7326540350914001,
+      "learning_rate": 0.00012774587436570141,
+      "loss": 0.22060327529907225,
+      "mean_token_accuracy": 0.9244944386184215,
+      "num_tokens": 4433050.0,
+      "step": 1900
+    },
+    {
+      "epoch": 4.7272727272727275,
+      "eval_entropy": 0.37579410852387896,
+      "eval_loss": 0.6031104922294617,
+      "eval_mean_token_accuracy": 0.8523658891749937,
+      "eval_num_tokens": 4433050.0,
+      "eval_runtime": 86.8475,
+      "eval_samples_per_second": 15.844,
+      "eval_steps_per_second": 1.98,
+      "step": 1900
+    },
+    {
+      "entropy": 0.28584762532263996,
+      "epoch": 4.777085927770859,
+      "grad_norm": 0.6692739725112915,
+      "learning_rate": 0.00012605559452841941,
+      "loss": 0.21919701099395753,
+      "mean_token_accuracy": 0.9259095810353756,
+      "num_tokens": 4483598.0,
+      "step": 1920
+    },
+    {
+      "epoch": 4.777085927770859,
+      "eval_entropy": 0.37441086370584575,
+      "eval_loss": 0.6009562611579895,
+      "eval_mean_token_accuracy": 0.8537772048351376,
+      "eval_num_tokens": 4483598.0,
+      "eval_runtime": 86.2697,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 1920
+    },
+    {
+      "entropy": 0.29914836324751376,
+      "epoch": 4.826899128268991,
+      "grad_norm": 0.8272984027862549,
+      "learning_rate": 0.00012435769088062695,
+      "loss": 0.22298691272735596,
+      "mean_token_accuracy": 0.9242447756230832,
+      "num_tokens": 4529201.0,
+      "step": 1940
+    },
+    {
+      "epoch": 4.826899128268991,
+      "eval_entropy": 0.37800539978021797,
+      "eval_loss": 0.5983167886734009,
+      "eval_mean_token_accuracy": 0.8530133397773255,
+      "eval_num_tokens": 4529201.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 1940
+    },
+    {
+      "entropy": 0.2881341265514493,
+      "epoch": 4.876712328767123,
+      "grad_norm": 0.8062415719032288,
+      "learning_rate": 0.00012265267548690605,
+      "loss": 0.21832451820373536,
+      "mean_token_accuracy": 0.9265582844614982,
+      "num_tokens": 4578167.0,
+      "step": 1960
+    },
+    {
+      "epoch": 4.876712328767123,
+      "eval_entropy": 0.3779863022094549,
+      "eval_loss": 0.5953031182289124,
+      "eval_mean_token_accuracy": 0.8528582415608472,
+      "eval_num_tokens": 4578167.0,
+      "eval_runtime": 86.392,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 1960
+    },
+    {
+      "entropy": 0.2998375216498971,
+      "epoch": 4.926525529265255,
+      "grad_norm": 0.8576586246490479,
+      "learning_rate": 0.00012094106255664416,
+      "loss": 0.22655115127563477,
+      "mean_token_accuracy": 0.9237113893032074,
+      "num_tokens": 4622316.0,
+      "step": 1980
+    },
+    {
+      "epoch": 4.926525529265255,
+      "eval_entropy": 0.3843588234726773,
+      "eval_loss": 0.5902820825576782,
+      "eval_mean_token_accuracy": 0.8534285481586013,
+      "eval_num_tokens": 4622316.0,
+      "eval_runtime": 86.5134,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 1980
+    },
+    {
+      "entropy": 0.2909785840660334,
+      "epoch": 4.976338729763388,
+      "grad_norm": 0.8082723021507263,
+      "learning_rate": 0.00011922336828895588,
+      "loss": 0.22233073711395263,
+      "mean_token_accuracy": 0.9257373474538326,
+      "num_tokens": 4670746.0,
+      "step": 2000
+    },
+    {
+      "epoch": 4.976338729763388,
+      "eval_entropy": 0.3669646157081737,
+      "eval_loss": 0.6064407229423523,
+      "eval_mean_token_accuracy": 0.8520209352637447,
+      "eval_num_tokens": 4670746.0,
+      "eval_runtime": 86.3523,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 1.992,
+      "step": 2000
+    },
+    {
+      "entropy": 0.2623713301160397,
+      "epoch": 5.024906600249066,
+      "grad_norm": 0.7109001874923706,
+      "learning_rate": 0.00011750011071700422,
+      "loss": 0.18533332347869874,
+      "mean_token_accuracy": 0.9396057151831113,
+      "num_tokens": 4716176.0,
+      "step": 2020
+    },
+    {
+      "epoch": 5.024906600249066,
+      "eval_entropy": 0.32857280068619305,
+      "eval_loss": 0.67312091588974,
+      "eval_mean_token_accuracy": 0.851016377640325,
+      "eval_num_tokens": 4716176.0,
+      "eval_runtime": 86.7326,
+      "eval_samples_per_second": 15.865,
+      "eval_steps_per_second": 1.983,
+      "step": 2020
+    },
+    {
+      "entropy": 0.1995823234319687,
+      "epoch": 5.074719800747198,
+      "grad_norm": 0.7890192270278931,
+      "learning_rate": 0.00011577180955176883,
+      "loss": 0.13584012985229493,
+      "mean_token_accuracy": 0.9562988370656967,
+      "num_tokens": 4763269.0,
+      "step": 2040
+    },
+    {
+      "epoch": 5.074719800747198,
+      "eval_entropy": 0.34224537690711576,
+      "eval_loss": 0.6417204141616821,
+      "eval_mean_token_accuracy": 0.8505163937807083,
+      "eval_num_tokens": 4763269.0,
+      "eval_runtime": 87.0157,
+      "eval_samples_per_second": 15.813,
+      "eval_steps_per_second": 1.977,
+      "step": 2040
+    },
+    {
+      "entropy": 0.19570088442414998,
+      "epoch": 5.12453300124533,
+      "grad_norm": 0.7778685688972473,
+      "learning_rate": 0.00011403898602530831,
+      "loss": 0.13429765701293944,
+      "mean_token_accuracy": 0.9550560437142849,
+      "num_tokens": 4809047.0,
+      "step": 2060
+    },
+    {
+      "epoch": 5.12453300124533,
+      "eval_entropy": 0.32213093167127566,
+      "eval_loss": 0.6681181788444519,
+      "eval_mean_token_accuracy": 0.8512208888003993,
+      "eval_num_tokens": 4809047.0,
+      "eval_runtime": 87.0793,
+      "eval_samples_per_second": 15.802,
+      "eval_steps_per_second": 1.975,
+      "step": 2060
+    },
+    {
+      "entropy": 0.19016577154397965,
+      "epoch": 5.174346201743462,
+      "grad_norm": 0.7051737904548645,
+      "learning_rate": 0.00011230216273356379,
+      "loss": 0.1268880009651184,
+      "mean_token_accuracy": 0.9573852688074111,
+      "num_tokens": 4857965.0,
+      "step": 2080
+    },
+    {
+      "epoch": 5.174346201743462,
+      "eval_entropy": 0.3113570500944936,
+      "eval_loss": 0.6709703207015991,
+      "eval_mean_token_accuracy": 0.8516059614891229,
+      "eval_num_tokens": 4857965.0,
+      "eval_runtime": 86.1197,
+      "eval_samples_per_second": 15.978,
+      "eval_steps_per_second": 1.997,
+      "step": 2080
+    },
+    {
+      "entropy": 0.20261877700686454,
+      "epoch": 5.224159402241594,
+      "grad_norm": 0.865691065788269,
+      "learning_rate": 0.00011056186347875099,
+      "loss": 0.13354524374008178,
+      "mean_token_accuracy": 0.9550505176186561,
+      "num_tokens": 4905151.0,
+      "step": 2100
+    },
+    {
+      "epoch": 5.224159402241594,
+      "eval_entropy": 0.3161145485071249,
+      "eval_loss": 0.6627876162528992,
+      "eval_mean_token_accuracy": 0.8524294512909513,
+      "eval_num_tokens": 4905151.0,
+      "eval_runtime": 86.8299,
+      "eval_samples_per_second": 15.847,
+      "eval_steps_per_second": 1.981,
+      "step": 2100
+    },
+    {
+      "entropy": 0.19859553650021552,
+      "epoch": 5.273972602739726,
+      "grad_norm": 0.6281663775444031,
+      "learning_rate": 0.00010881861311138865,
+      "loss": 0.13732362985610963,
+      "mean_token_accuracy": 0.9553415283560753,
+      "num_tokens": 4947072.0,
+      "step": 2120
+    },
+    {
+      "epoch": 5.273972602739726,
+      "eval_entropy": 0.30717471813739733,
+      "eval_loss": 0.6802853345870972,
+      "eval_mean_token_accuracy": 0.8511287407126538,
+      "eval_num_tokens": 4947072.0,
+      "eval_runtime": 86.5104,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2120
+    },
+    {
+      "entropy": 0.20874214079231024,
+      "epoch": 5.323785803237858,
+      "grad_norm": 0.8394920229911804,
+      "learning_rate": 0.00010707293737201094,
+      "loss": 0.13803558349609374,
+      "mean_token_accuracy": 0.9537043467164039,
+      "num_tokens": 4991974.0,
+      "step": 2140
+    },
+    {
+      "epoch": 5.323785803237858,
+      "eval_entropy": 0.3247084985812043,
+      "eval_loss": 0.6728973984718323,
+      "eval_mean_token_accuracy": 0.850993697033372,
+      "eval_num_tokens": 4991974.0,
+      "eval_runtime": 86.8565,
+      "eval_samples_per_second": 15.842,
+      "eval_steps_per_second": 1.98,
+      "step": 2140
+    },
+    {
+      "entropy": 0.20558980405330657,
+      "epoch": 5.37359900373599,
+      "grad_norm": 1.012776494026184,
+      "learning_rate": 0.00010532536273261094,
+      "loss": 0.14250640869140624,
+      "mean_token_accuracy": 0.9515879444777966,
+      "num_tokens": 5038400.0,
+      "step": 2160
+    },
+    {
+      "epoch": 5.37359900373599,
+      "eval_entropy": 0.33171969920743344,
+      "eval_loss": 0.6596540212631226,
+      "eval_mean_token_accuracy": 0.8504481239374294,
+      "eval_num_tokens": 5038400.0,
+      "eval_runtime": 86.8051,
+      "eval_samples_per_second": 15.852,
+      "eval_steps_per_second": 1.981,
+      "step": 2160
+    },
+    {
+      "entropy": 0.2098731081932783,
+      "epoch": 5.423412204234122,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00010357641623786434,
+      "loss": 0.13743176460266113,
+      "mean_token_accuracy": 0.9536496847867966,
+      "num_tokens": 5083166.0,
+      "step": 2180
+    },
+    {
+      "epoch": 5.423412204234122,
+      "eval_entropy": 0.31884251110428985,
+      "eval_loss": 0.6732954382896423,
+      "eval_mean_token_accuracy": 0.8506354866332786,
+      "eval_num_tokens": 5083166.0,
+      "eval_runtime": 86.8776,
+      "eval_samples_per_second": 15.838,
+      "eval_steps_per_second": 1.98,
+      "step": 2180
+    },
+    {
+      "entropy": 0.19826202262192966,
+      "epoch": 5.473225404732254,
+      "grad_norm": 0.6710811853408813,
+      "learning_rate": 0.00010182662534617945,
+      "loss": 0.13435200452804566,
+      "mean_token_accuracy": 0.9536436356604099,
+      "num_tokens": 5135109.0,
+      "step": 2200
+    },
+    {
+      "epoch": 5.473225404732254,
+      "eval_entropy": 0.3216065409225087,
+      "eval_loss": 0.6816763877868652,
+      "eval_mean_token_accuracy": 0.8507848818634831,
+      "eval_num_tokens": 5135109.0,
+      "eval_runtime": 86.6332,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2200
+    },
+    {
+      "entropy": 0.21997068710625173,
+      "epoch": 5.523038605230386,
+      "grad_norm": 0.757554292678833,
+      "learning_rate": 0.00010007651777062331,
+      "loss": 0.14061254262924194,
+      "mean_token_accuracy": 0.9530755899846554,
+      "num_tokens": 5180138.0,
+      "step": 2220
+    },
+    {
+      "epoch": 5.523038605230386,
+      "eval_entropy": 0.30639506633891617,
+      "eval_loss": 0.6736658811569214,
+      "eval_mean_token_accuracy": 0.8529234050318252,
+      "eval_num_tokens": 5180138.0,
+      "eval_runtime": 86.897,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 2220
+    },
+    {
+      "entropy": 0.20253944052383305,
+      "epoch": 5.572851805728518,
+      "grad_norm": 0.9304871559143066,
+      "learning_rate": 9.832662131977049e-05,
+      "loss": 0.1367262840270996,
+      "mean_token_accuracy": 0.9539065480232238,
+      "num_tokens": 5226864.0,
+      "step": 2240
+    },
+    {
+      "epoch": 5.572851805728518,
+      "eval_entropy": 0.3210238650614439,
+      "eval_loss": 0.6716064214706421,
+      "eval_mean_token_accuracy": 0.8515368752008261,
+      "eval_num_tokens": 5226864.0,
+      "eval_runtime": 86.6072,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 2240
+    },
+    {
+      "entropy": 0.204449025914073,
+      "epoch": 5.62266500622665,
+      "grad_norm": 0.5895705223083496,
+      "learning_rate": 9.657746373852313e-05,
+      "loss": 0.13733654022216796,
+      "mean_token_accuracy": 0.9532442130148411,
+      "num_tokens": 5276602.0,
+      "step": 2260
+    },
+    {
+      "epoch": 5.62266500622665,
+      "eval_entropy": 0.309385709724454,
+      "eval_loss": 0.6804041266441345,
+      "eval_mean_token_accuracy": 0.8522362411022186,
+      "eval_num_tokens": 5276602.0,
+      "eval_runtime": 86.5785,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 2260
+    },
+    {
+      "entropy": 0.1969501075334847,
+      "epoch": 5.672478206724782,
+      "grad_norm": 0.6863011121749878,
+      "learning_rate": 9.482957254895055e-05,
+      "loss": 0.13382177352905272,
+      "mean_token_accuracy": 0.9547011263668537,
+      "num_tokens": 5323718.0,
+      "step": 2280
+    },
+    {
+      "epoch": 5.672478206724782,
+      "eval_entropy": 0.31496484153145965,
+      "eval_loss": 0.6663601994514465,
+      "eval_mean_token_accuracy": 0.8531656691501307,
+      "eval_num_tokens": 5323718.0,
+      "eval_runtime": 86.688,
+      "eval_samples_per_second": 15.873,
+      "eval_steps_per_second": 1.984,
+      "step": 2280
+    },
+    {
+      "entropy": 0.2036854237318039,
+      "epoch": 5.722291407222914,
+      "grad_norm": 0.6446020603179932,
+      "learning_rate": 9.308347489119546e-05,
+      "loss": 0.1378130555152893,
+      "mean_token_accuracy": 0.9532770089805126,
+      "num_tokens": 5369039.0,
+      "step": 2300
+    },
+    {
+      "epoch": 5.722291407222914,
+      "eval_entropy": 0.3208180312500444,
+      "eval_loss": 0.6747905611991882,
+      "eval_mean_token_accuracy": 0.850395321152931,
+      "eval_num_tokens": 5369039.0,
+      "eval_runtime": 86.4375,
+      "eval_samples_per_second": 15.919,
+      "eval_steps_per_second": 1.99,
+      "step": 2300
+    },
+    {
+      "entropy": 0.20204954408109188,
+      "epoch": 5.772104607721046,
+      "grad_norm": 0.7637709379196167,
+      "learning_rate": 9.133969736449589e-05,
+      "loss": 0.13494755029678346,
+      "mean_token_accuracy": 0.9532943479716778,
+      "num_tokens": 5417029.0,
+      "step": 2320
+    },
+    {
+      "epoch": 5.772104607721046,
+      "eval_entropy": 0.3246596093441165,
+      "eval_loss": 0.6645228862762451,
+      "eval_mean_token_accuracy": 0.8514835512222245,
+      "eval_num_tokens": 5417029.0,
+      "eval_runtime": 86.251,
+      "eval_samples_per_second": 15.953,
+      "eval_steps_per_second": 1.994,
+      "step": 2320
+    },
+    {
+      "entropy": 0.21173860989511012,
+      "epoch": 5.821917808219178,
+      "grad_norm": 0.8394470810890198,
+      "learning_rate": 8.959876586837002e-05,
+      "loss": 0.1376061797142029,
+      "mean_token_accuracy": 0.9534688055515289,
+      "num_tokens": 5461671.0,
+      "step": 2340
+    },
+    {
+      "epoch": 5.821917808219178,
+      "eval_entropy": 0.3140063138548718,
+      "eval_loss": 0.6684557199478149,
+      "eval_mean_token_accuracy": 0.8525225107753,
+      "eval_num_tokens": 5461671.0,
+      "eval_runtime": 86.4656,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 2340
+    },
+    {
+      "entropy": 0.1935113413259387,
+      "epoch": 5.87173100871731,
+      "grad_norm": 0.6055243611335754,
+      "learning_rate": 8.786120544401189e-05,
+      "loss": 0.1318501591682434,
+      "mean_token_accuracy": 0.955411896854639,
+      "num_tokens": 5509312.0,
+      "step": 2360
+    },
+    {
+      "epoch": 5.87173100871731,
+      "eval_entropy": 0.31226280764784925,
+      "eval_loss": 0.6752243041992188,
+      "eval_mean_token_accuracy": 0.8520811529353608,
+      "eval_num_tokens": 5509312.0,
+      "eval_runtime": 86.4078,
+      "eval_samples_per_second": 15.924,
+      "eval_steps_per_second": 1.991,
+      "step": 2360
+    },
+    {
+      "entropy": 0.2034847954288125,
+      "epoch": 5.921544209215442,
+      "grad_norm": 0.6410462260246277,
+      "learning_rate": 8.612754011594654e-05,
+      "loss": 0.1375146269798279,
+      "mean_token_accuracy": 0.9542048372328281,
+      "num_tokens": 5558196.0,
+      "step": 2380
+    },
+    {
+      "epoch": 5.921544209215442,
+      "eval_entropy": 0.3186545864960482,
+      "eval_loss": 0.6681073904037476,
+      "eval_mean_token_accuracy": 0.851885937674101,
+      "eval_num_tokens": 5558196.0,
+      "eval_runtime": 86.4004,
+      "eval_samples_per_second": 15.926,
+      "eval_steps_per_second": 1.991,
+      "step": 2380
+    },
+    {
+      "entropy": 0.19530893191695214,
+      "epoch": 5.971357409713574,
+      "grad_norm": 0.7243332266807556,
+      "learning_rate": 8.439829273399083e-05,
+      "loss": 0.13524004220962524,
+      "mean_token_accuracy": 0.9542504735291004,
+      "num_tokens": 5605400.0,
+      "step": 2400
+    },
+    {
+      "epoch": 5.971357409713574,
+      "eval_entropy": 0.31358762431976406,
+      "eval_loss": 0.6792373061180115,
+      "eval_mean_token_accuracy": 0.8525331470855447,
+      "eval_num_tokens": 5605400.0,
+      "eval_runtime": 86.4904,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 2400
+    },
+    {
+      "entropy": 0.20260333069241965,
+      "epoch": 6.019925280199253,
+      "grad_norm": 0.43665260076522827,
+      "learning_rate": 8.267398481556944e-05,
+      "loss": 0.1221657395362854,
+      "mean_token_accuracy": 0.958129896567418,
+      "num_tokens": 5648968.0,
+      "step": 2420
+    },
+    {
+      "epoch": 6.019925280199253,
+      "eval_entropy": 0.29224625195181647,
+      "eval_loss": 0.7269865274429321,
+      "eval_mean_token_accuracy": 0.8504481551259063,
+      "eval_num_tokens": 5648968.0,
+      "eval_runtime": 86.7121,
+      "eval_samples_per_second": 15.869,
+      "eval_steps_per_second": 1.984,
+      "step": 2420
+    },
+    {
+      "entropy": 0.1478068961761892,
+      "epoch": 6.069738480697385,
+      "grad_norm": 0.5789358019828796,
+      "learning_rate": 8.095513638843222e-05,
+      "loss": 0.08725568056106567,
+      "mean_token_accuracy": 0.9692194186151027,
+      "num_tokens": 5697633.0,
+      "step": 2440
+    },
+    {
+      "epoch": 6.069738480697385,
+      "eval_entropy": 0.2874765749587569,
+      "eval_loss": 0.7498268485069275,
+      "eval_mean_token_accuracy": 0.8493660684934882,
+      "eval_num_tokens": 5697633.0,
+      "eval_runtime": 85.9922,
+      "eval_samples_per_second": 16.001,
+      "eval_steps_per_second": 2.0,
+      "step": 2440
+    },
+    {
+      "entropy": 0.1555120893754065,
+      "epoch": 6.119551681195516,
+      "grad_norm": 0.4853467643260956,
+      "learning_rate": 7.924226583382094e-05,
+      "loss": 0.09054291844367982,
+      "mean_token_accuracy": 0.9687649801373481,
+      "num_tokens": 5745090.0,
+      "step": 2460
+    },
+    {
+      "epoch": 6.119551681195516,
+      "eval_entropy": 0.28221732412659845,
+      "eval_loss": 0.7451956272125244,
+      "eval_mean_token_accuracy": 0.8512702442185823,
+      "eval_num_tokens": 5745090.0,
+      "eval_runtime": 87.7186,
+      "eval_samples_per_second": 15.687,
+      "eval_steps_per_second": 1.961,
+      "step": 2460
+    },
+    {
+      "entropy": 0.15636382224038242,
+      "epoch": 6.169364881693649,
+      "grad_norm": 0.484597772359848,
+      "learning_rate": 7.753588973013247e-05,
+      "loss": 0.08987976908683777,
+      "mean_token_accuracy": 0.9688864842057228,
+      "num_tokens": 5790333.0,
+      "step": 2480
+    },
+    {
+      "epoch": 6.169364881693649,
+      "eval_entropy": 0.2812082867116429,
+      "eval_loss": 0.7391297817230225,
+      "eval_mean_token_accuracy": 0.8511726624050806,
+      "eval_num_tokens": 5790333.0,
+      "eval_runtime": 86.2049,
+      "eval_samples_per_second": 15.962,
+      "eval_steps_per_second": 1.995,
+      "step": 2480
+    },
+    {
+      "entropy": 0.1488968376070261,
+      "epoch": 6.219178082191781,
+      "grad_norm": 0.29219967126846313,
+      "learning_rate": 7.583652269712595e-05,
+      "loss": 0.08845279812812805,
+      "mean_token_accuracy": 0.9708508215844631,
+      "num_tokens": 5837321.0,
+      "step": 2500
+    },
+    {
+      "epoch": 6.219178082191781,
+      "eval_entropy": 0.27179383755076764,
+      "eval_loss": 0.760672390460968,
+      "eval_mean_token_accuracy": 0.8512875420409579,
+      "eval_num_tokens": 5837321.0,
+      "eval_runtime": 86.3888,
+      "eval_samples_per_second": 15.928,
+      "eval_steps_per_second": 1.991,
+      "step": 2500
+    },
+    {
+      "entropy": 0.14728794833645226,
+      "epoch": 6.268991282689913,
+      "grad_norm": 0.6203164458274841,
+      "learning_rate": 7.414467724072035e-05,
+      "loss": 0.08744659423828124,
+      "mean_token_accuracy": 0.9708136722445488,
+      "num_tokens": 5884398.0,
+      "step": 2520
+    },
+    {
+      "epoch": 6.268991282689913,
+      "eval_entropy": 0.2740729117636071,
+      "eval_loss": 0.7610964179039001,
+      "eval_mean_token_accuracy": 0.8515192301467408,
+      "eval_num_tokens": 5884398.0,
+      "eval_runtime": 86.2266,
+      "eval_samples_per_second": 15.958,
+      "eval_steps_per_second": 1.995,
+      "step": 2520
+    },
+    {
+      "entropy": 0.15153271816670894,
+      "epoch": 6.318804483188045,
+      "grad_norm": 0.5843842625617981,
+      "learning_rate": 7.246086359842966e-05,
+      "loss": 0.09105724096298218,
+      "mean_token_accuracy": 0.9692696832120419,
+      "num_tokens": 5929876.0,
+      "step": 2540
+    },
+    {
+      "epoch": 6.318804483188045,
+      "eval_entropy": 0.27398888333592303,
+      "eval_loss": 0.746191680431366,
+      "eval_mean_token_accuracy": 0.8532331041818442,
+      "eval_num_tokens": 5929876.0,
+      "eval_runtime": 86.2203,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2540
+    },
+    {
+      "entropy": 0.14994749650359154,
+      "epoch": 6.3686176836861765,
+      "grad_norm": 0.6314879655838013,
+      "learning_rate": 7.078558958548248e-05,
+      "loss": 0.08776435852050782,
+      "mean_token_accuracy": 0.9698216117918491,
+      "num_tokens": 5978490.0,
+      "step": 2560
+    },
+    {
+      "epoch": 6.3686176836861765,
+      "eval_entropy": 0.28109629790103713,
+      "eval_loss": 0.7394832968711853,
+      "eval_mean_token_accuracy": 0.8526893916518189,
+      "eval_num_tokens": 5978490.0,
+      "eval_runtime": 86.5096,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 2560
+    },
+    {
+      "entropy": 0.1454140623100102,
+      "epoch": 6.418430884184309,
+      "grad_norm": 0.6952065229415894,
+      "learning_rate": 6.911936044167146e-05,
+      "loss": 0.08621888756752014,
+      "mean_token_accuracy": 0.9713154613971711,
+      "num_tokens": 6025380.0,
+      "step": 2580
+    },
+    {
+      "epoch": 6.418430884184309,
+      "eval_entropy": 0.2741171905640946,
+      "eval_loss": 0.7543420195579529,
+      "eval_mean_token_accuracy": 0.8530452431634415,
+      "eval_num_tokens": 6025380.0,
+      "eval_runtime": 86.2748,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 2580
+    },
+    {
+      "entropy": 0.1498984983190894,
+      "epoch": 6.468244084682441,
+      "grad_norm": 0.48904216289520264,
+      "learning_rate": 6.746267867898023e-05,
+      "loss": 0.08652682900428772,
+      "mean_token_accuracy": 0.9702693298459053,
+      "num_tokens": 6073349.0,
+      "step": 2600
+    },
+    {
+      "epoch": 6.468244084682441,
+      "eval_entropy": 0.2762772881880749,
+      "eval_loss": 0.7408088445663452,
+      "eval_mean_token_accuracy": 0.8531355317248854,
+      "eval_num_tokens": 6073349.0,
+      "eval_runtime": 86.3263,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 2600
+    },
+    {
+      "entropy": 0.14868942834436893,
+      "epoch": 6.518057285180573,
+      "grad_norm": 0.5196901559829712,
+      "learning_rate": 6.581604393003236e-05,
+      "loss": 0.09130032062530517,
+      "mean_token_accuracy": 0.9702563397586346,
+      "num_tokens": 6118148.0,
+      "step": 2620
+    },
+    {
+      "epoch": 6.518057285180573,
+      "eval_entropy": 0.26881319327756414,
+      "eval_loss": 0.7546963095664978,
+      "eval_mean_token_accuracy": 0.852999032236809,
+      "eval_num_tokens": 6118148.0,
+      "eval_runtime": 86.3644,
+      "eval_samples_per_second": 15.932,
+      "eval_steps_per_second": 1.992,
+      "step": 2620
+    },
+    {
+      "entropy": 0.14665521653369068,
+      "epoch": 6.567870485678705,
+      "grad_norm": 0.5889317393302917,
+      "learning_rate": 6.417995279740927e-05,
+      "loss": 0.09003536105155945,
+      "mean_token_accuracy": 0.9698039405047894,
+      "num_tokens": 6165523.0,
+      "step": 2640
+    },
+    {
+      "epoch": 6.567870485678705,
+      "eval_entropy": 0.2733227195608061,
+      "eval_loss": 0.7500578165054321,
+      "eval_mean_token_accuracy": 0.8526958410823068,
+      "eval_num_tokens": 6165523.0,
+      "eval_runtime": 86.257,
+      "eval_samples_per_second": 15.952,
+      "eval_steps_per_second": 1.994,
+      "step": 2640
+    },
+    {
+      "entropy": 0.14399882415309548,
+      "epoch": 6.617683686176837,
+      "grad_norm": 0.6247589588165283,
+      "learning_rate": 6.25548987038814e-05,
+      "loss": 0.08689293265342712,
+      "mean_token_accuracy": 0.971392173320055,
+      "num_tokens": 6216232.0,
+      "step": 2660
+    },
+    {
+      "epoch": 6.617683686176837,
+      "eval_entropy": 0.27525238363548765,
+      "eval_loss": 0.7506860494613647,
+      "eval_mean_token_accuracy": 0.8529516337222831,
+      "eval_num_tokens": 6216232.0,
+      "eval_runtime": 86.1844,
+      "eval_samples_per_second": 15.966,
+      "eval_steps_per_second": 1.996,
+      "step": 2660
+    },
+    {
+      "entropy": 0.14995302110910416,
+      "epoch": 6.667496886674969,
+      "grad_norm": 0.6068060994148254,
+      "learning_rate": 6.0941371743598745e-05,
+      "loss": 0.08885984420776367,
+      "mean_token_accuracy": 0.9696493484079838,
+      "num_tokens": 6264807.0,
+      "step": 2680
+    },
+    {
+      "epoch": 6.667496886674969,
+      "eval_entropy": 0.2835908398725266,
+      "eval_loss": 0.7458326816558838,
+      "eval_mean_token_accuracy": 0.8517367382382237,
+      "eval_num_tokens": 6264807.0,
+      "eval_runtime": 86.66,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 2680
+    },
+    {
+      "entropy": 0.16357141444459558,
+      "epoch": 6.717310087173101,
+      "grad_norm": 0.6181513667106628,
+      "learning_rate": 5.933985853428513e-05,
+      "loss": 0.09438056945800781,
+      "mean_token_accuracy": 0.9676273122429848,
+      "num_tokens": 6308819.0,
+      "step": 2700
+    },
+    {
+      "epoch": 6.717310087173101,
+      "eval_entropy": 0.2713852625253589,
+      "eval_loss": 0.7483112215995789,
+      "eval_mean_token_accuracy": 0.8531875988078672,
+      "eval_num_tokens": 6308819.0,
+      "eval_runtime": 86.385,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 2700
+    },
+    {
+      "entropy": 0.1630424507893622,
+      "epoch": 6.767123287671232,
+      "grad_norm": 0.46826601028442383,
+      "learning_rate": 5.775084207048101e-05,
+      "loss": 0.09234987497329712,
+      "mean_token_accuracy": 0.9669895999133586,
+      "num_tokens": 6352548.0,
+      "step": 2720
+    },
+    {
+      "epoch": 6.767123287671232,
+      "eval_entropy": 0.26874022653629615,
+      "eval_loss": 0.7569906711578369,
+      "eval_mean_token_accuracy": 0.852455192180567,
+      "eval_num_tokens": 6352548.0,
+      "eval_runtime": 86.196,
+      "eval_samples_per_second": 15.964,
+      "eval_steps_per_second": 1.995,
+      "step": 2720
+    },
+    {
+      "entropy": 0.14564633751288056,
+      "epoch": 6.816936488169365,
+      "grad_norm": 0.3920297920703888,
+      "learning_rate": 5.6174801577878795e-05,
+      "loss": 0.08979449868202209,
+      "mean_token_accuracy": 0.9692104034125805,
+      "num_tokens": 6397720.0,
+      "step": 2740
+    },
+    {
+      "epoch": 6.816936488169365,
+      "eval_entropy": 0.26479411177163903,
+      "eval_loss": 0.7598947882652283,
+      "eval_mean_token_accuracy": 0.8536758384732313,
+      "eval_num_tokens": 6397720.0,
+      "eval_runtime": 86.1023,
+      "eval_samples_per_second": 15.981,
+      "eval_steps_per_second": 1.998,
+      "step": 2740
+    },
+    {
+      "entropy": 0.16201716642826797,
+      "epoch": 6.866749688667497,
+      "grad_norm": 0.496540367603302,
+      "learning_rate": 5.4612212368794904e-05,
+      "loss": 0.09179213643074036,
+      "mean_token_accuracy": 0.9680421657860279,
+      "num_tokens": 6442821.0,
+      "step": 2760
+    },
+    {
+      "epoch": 6.866749688667497,
+      "eval_entropy": 0.27176168924847316,
+      "eval_loss": 0.756536602973938,
+      "eval_mean_token_accuracy": 0.8528874596191007,
+      "eval_num_tokens": 6442821.0,
+      "eval_runtime": 86.1371,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2760
+    },
+    {
+      "entropy": 0.13891819408163428,
+      "epoch": 6.916562889165629,
+      "grad_norm": 0.5002485513687134,
+      "learning_rate": 5.3063545698822274e-05,
+      "loss": 0.08300527930259705,
+      "mean_token_accuracy": 0.971545273065567,
+      "num_tokens": 6494118.0,
+      "step": 2780
+    },
+    {
+      "epoch": 6.916562889165629,
+      "eval_entropy": 0.26986357344444406,
+      "eval_loss": 0.7550951242446899,
+      "eval_mean_token_accuracy": 0.8532527044068935,
+      "eval_num_tokens": 6494118.0,
+      "eval_runtime": 86.642,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 2780
+    },
+    {
+      "entropy": 0.14777150927111507,
+      "epoch": 6.966376089663761,
+      "grad_norm": 0.35520732402801514,
+      "learning_rate": 5.152926862470571e-05,
+      "loss": 0.08964785933494568,
+      "mean_token_accuracy": 0.9702881813049317,
+      "num_tokens": 6540175.0,
+      "step": 2800
+    },
+    {
+      "epoch": 6.966376089663761,
+      "eval_entropy": 0.26697341023489485,
+      "eval_loss": 0.753479540348053,
+      "eval_mean_token_accuracy": 0.8532822423203047,
+      "eval_num_tokens": 6540175.0,
+      "eval_runtime": 86.6313,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 2800
+    },
+    {
+      "entropy": 0.16215767501256403,
+      "epoch": 7.01494396014944,
+      "grad_norm": 0.41724473237991333,
+      "learning_rate": 5.0009843863484395e-05,
+      "loss": 0.0857824981212616,
+      "mean_token_accuracy": 0.9699622369729556,
+      "num_tokens": 6583767.0,
+      "step": 2820
+    },
+    {
+      "epoch": 7.01494396014944,
+      "eval_entropy": 0.255824513473483,
+      "eval_loss": 0.773139238357544,
+      "eval_mean_token_accuracy": 0.8542473659265873,
+      "eval_num_tokens": 6583767.0,
+      "eval_runtime": 86.2663,
+      "eval_samples_per_second": 15.951,
+      "eval_steps_per_second": 1.994,
+      "step": 2820
+    },
+    {
+      "entropy": 0.13403823282569646,
+      "epoch": 7.064757160647572,
+      "grad_norm": 0.5895716547966003,
+      "learning_rate": 4.8505729652942433e-05,
+      "loss": 0.07477729320526123,
+      "mean_token_accuracy": 0.9744169726967812,
+      "num_tokens": 6627672.0,
+      "step": 2840
+    },
+    {
+      "epoch": 7.064757160647572,
+      "eval_entropy": 0.24803267142107321,
+      "eval_loss": 0.8059361577033997,
+      "eval_mean_token_accuracy": 0.8533402220454327,
+      "eval_num_tokens": 6627672.0,
+      "eval_runtime": 86.2105,
+      "eval_samples_per_second": 15.961,
+      "eval_steps_per_second": 1.995,
+      "step": 2840
+    },
+    {
+      "entropy": 0.12771664513275027,
+      "epoch": 7.114570361145703,
+      "grad_norm": 0.22397534549236298,
+      "learning_rate": 4.7017379613410626e-05,
+      "loss": 0.06963326930999755,
+      "mean_token_accuracy": 0.9753161281347275,
+      "num_tokens": 6672946.0,
+      "step": 2860
+    },
+    {
+      "epoch": 7.114570361145703,
+      "eval_entropy": 0.2530888926324456,
+      "eval_loss": 0.809047520160675,
+      "eval_mean_token_accuracy": 0.8522793931323428,
+      "eval_num_tokens": 6672946.0,
+      "eval_runtime": 86.219,
+      "eval_samples_per_second": 15.959,
+      "eval_steps_per_second": 1.995,
+      "step": 2860
+    },
+    {
+      "entropy": 0.12172898761928082,
+      "epoch": 7.164383561643835,
+      "grad_norm": 0.5140368342399597,
+      "learning_rate": 4.554524261096081e-05,
+      "loss": 0.06878940463066101,
+      "mean_token_accuracy": 0.9762930549681187,
+      "num_tokens": 6718582.0,
+      "step": 2880
+    },
+    {
+      "epoch": 7.164383561643835,
+      "eval_entropy": 0.2459391265414482,
+      "eval_loss": 0.8392541408538818,
+      "eval_mean_token_accuracy": 0.8528771050447641,
+      "eval_num_tokens": 6718582.0,
+      "eval_runtime": 85.9805,
+      "eval_samples_per_second": 16.004,
+      "eval_steps_per_second": 2.0,
+      "step": 2880
+    },
+    {
+      "entropy": 0.12581687197089195,
+      "epoch": 7.214196762141968,
+      "grad_norm": 0.28926676511764526,
+      "learning_rate": 4.408976262203404e-05,
+      "loss": 0.07077260017395019,
+      "mean_token_accuracy": 0.975715234130621,
+      "num_tokens": 6764387.0,
+      "step": 2900
+    },
+    {
+      "epoch": 7.214196762141968,
+      "eval_entropy": 0.24933571585042533,
+      "eval_loss": 0.8336001634597778,
+      "eval_mean_token_accuracy": 0.8527654329704684,
+      "eval_num_tokens": 6764387.0,
+      "eval_runtime": 85.2853,
+      "eval_samples_per_second": 16.134,
+      "eval_steps_per_second": 2.017,
+      "step": 2900
+    },
+    {
+      "entropy": 0.13015523147769273,
+      "epoch": 7.2640099626401,
+      "grad_norm": 0.31048327684402466,
+      "learning_rate": 4.2651378599543286e-05,
+      "loss": 0.07249900102615356,
+      "mean_token_accuracy": 0.9749723896384239,
+      "num_tokens": 6809678.0,
+      "step": 2920
+    },
+    {
+      "epoch": 7.2640099626401,
+      "eval_entropy": 0.2481014168886251,
+      "eval_loss": 0.8247711062431335,
+      "eval_mean_token_accuracy": 0.8533733261879101,
+      "eval_num_tokens": 6809678.0,
+      "eval_runtime": 86.1337,
+      "eval_samples_per_second": 15.975,
+      "eval_steps_per_second": 1.997,
+      "step": 2920
+    },
+    {
+      "entropy": 0.11947155892848968,
+      "epoch": 7.313823163138232,
+      "grad_norm": 0.3158417046070099,
+      "learning_rate": 4.123052434049112e-05,
+      "loss": 0.06591642498970032,
+      "mean_token_accuracy": 0.9769096739590168,
+      "num_tokens": 6860750.0,
+      "step": 2940
+    },
+    {
+      "epoch": 7.313823163138232,
+      "eval_entropy": 0.24779942059932752,
+      "eval_loss": 0.8287138938903809,
+      "eval_mean_token_accuracy": 0.8529684456974961,
+      "eval_num_tokens": 6860750.0,
+      "eval_runtime": 86.1438,
+      "eval_samples_per_second": 15.973,
+      "eval_steps_per_second": 1.997,
+      "step": 2940
+    },
+    {
+      "entropy": 0.13140737023204566,
+      "epoch": 7.363636363636363,
+      "grad_norm": 0.5523769855499268,
+      "learning_rate": 3.9827628355142654e-05,
+      "loss": 0.0709549367427826,
+      "mean_token_accuracy": 0.974937304854393,
+      "num_tokens": 6907092.0,
+      "step": 2960
+    },
+    {
+      "epoch": 7.363636363636363,
+      "eval_entropy": 0.24850021077449932,
+      "eval_loss": 0.8252517580986023,
+      "eval_mean_token_accuracy": 0.8539477028819018,
+      "eval_num_tokens": 6907092.0,
+      "eval_runtime": 86.4886,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 2960
+    },
+    {
+      "entropy": 0.13003513137809933,
+      "epoch": 7.4134495641344955,
+      "grad_norm": 0.4664781391620636,
+      "learning_rate": 3.844311373779252e-05,
+      "loss": 0.07174457907676697,
+      "mean_token_accuracy": 0.9750649809837342,
+      "num_tokens": 6952700.0,
+      "step": 2980
+    },
+    {
+      "epoch": 7.4134495641344955,
+      "eval_entropy": 0.24727656943506973,
+      "eval_loss": 0.8217361569404602,
+      "eval_mean_token_accuracy": 0.8543338027111319,
+      "eval_num_tokens": 6952700.0,
+      "eval_runtime": 86.2981,
+      "eval_samples_per_second": 15.945,
+      "eval_steps_per_second": 1.993,
+      "step": 2980
+    },
+    {
+      "entropy": 0.13791935429908336,
+      "epoch": 7.463262764632628,
+      "grad_norm": 0.30846619606018066,
+      "learning_rate": 3.7077398039165516e-05,
+      "loss": 0.07441290616989135,
+      "mean_token_accuracy": 0.9730765089392662,
+      "num_tokens": 6996430.0,
+      "step": 3000
+    },
+    {
+      "epoch": 7.463262764632628,
+      "eval_entropy": 0.25024019753517107,
+      "eval_loss": 0.8234542012214661,
+      "eval_mean_token_accuracy": 0.8529301455547643,
+      "eval_num_tokens": 6996430.0,
+      "eval_runtime": 86.1868,
+      "eval_samples_per_second": 15.965,
+      "eval_steps_per_second": 1.996,
+      "step": 3000
+    },
+    {
+      "entropy": 0.1354162965901196,
+      "epoch": 7.51307596513076,
+      "grad_norm": 0.30689817667007446,
+      "learning_rate": 3.573089314048897e-05,
+      "loss": 0.07313602566719055,
+      "mean_token_accuracy": 0.9747377909719944,
+      "num_tokens": 7042038.0,
+      "step": 3020
+    },
+    {
+      "epoch": 7.51307596513076,
+      "eval_entropy": 0.24558737484175105,
+      "eval_loss": 0.8214292526245117,
+      "eval_mean_token_accuracy": 0.8544686381206956,
+      "eval_num_tokens": 7042038.0,
+      "eval_runtime": 86.406,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3020
+    },
+    {
+      "entropy": 0.12779442030005156,
+      "epoch": 7.562889165628892,
+      "grad_norm": 0.4416750967502594,
+      "learning_rate": 3.4404005129274905e-05,
+      "loss": 0.07139642834663391,
+      "mean_token_accuracy": 0.9753781862556934,
+      "num_tokens": 7089390.0,
+      "step": 3040
+    },
+    {
+      "epoch": 7.562889165628892,
+      "eval_entropy": 0.24580405817128892,
+      "eval_loss": 0.8302819132804871,
+      "eval_mean_token_accuracy": 0.8540662842434507,
+      "eval_num_tokens": 7089390.0,
+      "eval_runtime": 86.5957,
+      "eval_samples_per_second": 15.89,
+      "eval_steps_per_second": 1.986,
+      "step": 3040
+    },
+    {
+      "entropy": 0.12047688071615995,
+      "epoch": 7.6127023661270234,
+      "grad_norm": 0.5483342409133911,
+      "learning_rate": 3.309713417684963e-05,
+      "loss": 0.06730042099952697,
+      "mean_token_accuracy": 0.9768634103238583,
+      "num_tokens": 7138002.0,
+      "step": 3060
+    },
+    {
+      "epoch": 7.6127023661270234,
+      "eval_entropy": 0.24299402422336644,
+      "eval_loss": 0.831507682800293,
+      "eval_mean_token_accuracy": 0.8545246692590935,
+      "eval_num_tokens": 7138002.0,
+      "eval_runtime": 86.4733,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 3060
+    },
+    {
+      "entropy": 0.1223155280109495,
+      "epoch": 7.662515566625156,
+      "grad_norm": 0.23236846923828125,
+      "learning_rate": 3.181067441766772e-05,
+      "loss": 0.06945533752441406,
+      "mean_token_accuracy": 0.9759405627846718,
+      "num_tokens": 7185520.0,
+      "step": 3080
+    },
+    {
+      "epoch": 7.662515566625156,
+      "eval_entropy": 0.24285893584060114,
+      "eval_loss": 0.8383913040161133,
+      "eval_mean_token_accuracy": 0.8540910773499067,
+      "eval_num_tokens": 7185520.0,
+      "eval_runtime": 86.4055,
+      "eval_samples_per_second": 15.925,
+      "eval_steps_per_second": 1.991,
+      "step": 3080
+    },
+    {
+      "entropy": 0.12549724937416612,
+      "epoch": 7.712328767123288,
+      "grad_norm": 0.5445857048034668,
+      "learning_rate": 3.054501383044605e-05,
+      "loss": 0.07318683862686157,
+      "mean_token_accuracy": 0.9748943917453289,
+      "num_tokens": 7232170.0,
+      "step": 3100
+    },
+    {
+      "epoch": 7.712328767123288,
+      "eval_entropy": 0.24052567260209903,
+      "eval_loss": 0.8307085037231445,
+      "eval_mean_token_accuracy": 0.8547526040049487,
+      "eval_num_tokens": 7232170.0,
+      "eval_runtime": 86.627,
+      "eval_samples_per_second": 15.884,
+      "eval_steps_per_second": 1.986,
+      "step": 3100
+    },
+    {
+      "entropy": 0.11182957985438406,
+      "epoch": 7.76214196762142,
+      "grad_norm": 0.5338284969329834,
+      "learning_rate": 2.9300534121155e-05,
+      "loss": 0.06559510231018066,
+      "mean_token_accuracy": 0.9782855957746506,
+      "num_tokens": 7282846.0,
+      "step": 3120
+    },
+    {
+      "epoch": 7.76214196762142,
+      "eval_entropy": 0.2438461010365985,
+      "eval_loss": 0.8223569989204407,
+      "eval_mean_token_accuracy": 0.8548611496077028,
+      "eval_num_tokens": 7282846.0,
+      "eval_runtime": 86.6,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3120
+    },
+    {
+      "entropy": 0.12523467522114515,
+      "epoch": 7.811955168119551,
+      "grad_norm": 0.4262288808822632,
+      "learning_rate": 2.807761060790107e-05,
+      "loss": 0.06978838443756104,
+      "mean_token_accuracy": 0.9748328007757664,
+      "num_tokens": 7331518.0,
+      "step": 3140
+    },
+    {
+      "epoch": 7.811955168119551,
+      "eval_entropy": 0.2413516822595929,
+      "eval_loss": 0.8272029161453247,
+      "eval_mean_token_accuracy": 0.8549864254025525,
+      "eval_num_tokens": 7331518.0,
+      "eval_runtime": 86.6091,
+      "eval_samples_per_second": 15.887,
+      "eval_steps_per_second": 1.986,
+      "step": 3140
+    },
+    {
+      "entropy": 0.12787675098516046,
+      "epoch": 7.861768368617684,
+      "grad_norm": 0.4428657591342926,
+      "learning_rate": 2.687661210773607e-05,
+      "loss": 0.06889715790748596,
+      "mean_token_accuracy": 0.973549035936594,
+      "num_tokens": 7379056.0,
+      "step": 3160
+    },
+    {
+      "epoch": 7.861768368617684,
+      "eval_entropy": 0.24256615280065424,
+      "eval_loss": 0.8303911089897156,
+      "eval_mean_token_accuracy": 0.8550452936527341,
+      "eval_num_tokens": 7379056.0,
+      "eval_runtime": 86.637,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3160
+    },
+    {
+      "entropy": 0.12699329359456896,
+      "epoch": 7.911581569115816,
+      "grad_norm": 0.20887182652950287,
+      "learning_rate": 2.5697900825426814e-05,
+      "loss": 0.07120747566223144,
+      "mean_token_accuracy": 0.9752186723053455,
+      "num_tokens": 7424770.0,
+      "step": 3180
+    },
+    {
+      "epoch": 7.911581569115816,
+      "eval_entropy": 0.24418324786563253,
+      "eval_loss": 0.825463593006134,
+      "eval_mean_token_accuracy": 0.8547488301992416,
+      "eval_num_tokens": 7424770.0,
+      "eval_runtime": 86.4434,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 3180
+    },
+    {
+      "entropy": 0.13169807861559094,
+      "epoch": 7.961394769613948,
+      "grad_norm": 0.1804407387971878,
+      "learning_rate": 2.4541832244219308e-05,
+      "loss": 0.0711442768573761,
+      "mean_token_accuracy": 0.9736054383218289,
+      "num_tokens": 7470804.0,
+      "step": 3200
+    },
+    {
+      "epoch": 7.961394769613948,
+      "eval_entropy": 0.2437171469074349,
+      "eval_loss": 0.8287582993507385,
+      "eval_mean_token_accuracy": 0.8545196194981419,
+      "eval_num_tokens": 7470804.0,
+      "eval_runtime": 86.6395,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 3200
+    },
+    {
+      "entropy": 0.11447950548086411,
+      "epoch": 8.009962640099626,
+      "grad_norm": 0.11681001633405685,
+      "learning_rate": 2.3408755018629662e-05,
+      "loss": 0.0648022472858429,
+      "mean_token_accuracy": 0.9776463088316795,
+      "num_tokens": 7518110.0,
+      "step": 3220
+    },
+    {
+      "epoch": 8.009962640099626,
+      "eval_entropy": 0.2403542645448862,
+      "eval_loss": 0.8375888466835022,
+      "eval_mean_token_accuracy": 0.8546101745477942,
+      "eval_num_tokens": 7518110.0,
+      "eval_runtime": 86.6664,
+      "eval_samples_per_second": 15.877,
+      "eval_steps_per_second": 1.985,
+      "step": 3220
+    },
+    {
+      "entropy": 0.12123392843641341,
+      "epoch": 8.059775840597759,
+      "grad_norm": 0.14732785522937775,
+      "learning_rate": 2.229901086929465e-05,
+      "loss": 0.06151747703552246,
+      "mean_token_accuracy": 0.9774191610515117,
+      "num_tokens": 7565539.0,
+      "step": 3240
+    },
+    {
+      "epoch": 8.059775840597759,
+      "eval_entropy": 0.2306776445272357,
+      "eval_loss": 0.8743928670883179,
+      "eval_mean_token_accuracy": 0.8543573676846749,
+      "eval_num_tokens": 7565539.0,
+      "eval_runtime": 86.5684,
+      "eval_samples_per_second": 15.895,
+      "eval_steps_per_second": 1.987,
+      "step": 3240
+    },
+    {
+      "entropy": 0.1161969758104533,
+      "epoch": 8.10958904109589,
+      "grad_norm": 0.15050369501113892,
+      "learning_rate": 2.1212934479913717e-05,
+      "loss": 0.06055923104286194,
+      "mean_token_accuracy": 0.9778320029377937,
+      "num_tokens": 7612578.0,
+      "step": 3260
+    },
+    {
+      "epoch": 8.10958904109589,
+      "eval_entropy": 0.2284980574839337,
+      "eval_loss": 0.8921975493431091,
+      "eval_mean_token_accuracy": 0.853883276498595,
+      "eval_num_tokens": 7612578.0,
+      "eval_runtime": 86.5997,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3260
+    },
+    {
+      "entropy": 0.11528769815340638,
+      "epoch": 8.159402241594023,
+      "grad_norm": 0.13709086179733276,
+      "learning_rate": 2.0150853396312505e-05,
+      "loss": 0.060501879453659056,
+      "mean_token_accuracy": 0.9761978469789028,
+      "num_tokens": 7660157.0,
+      "step": 3280
+    },
+    {
+      "epoch": 8.159402241594023,
+      "eval_entropy": 0.23356944929028667,
+      "eval_loss": 0.8960044384002686,
+      "eval_mean_token_accuracy": 0.8533267111972321,
+      "eval_num_tokens": 7660157.0,
+      "eval_runtime": 86.1662,
+      "eval_samples_per_second": 15.969,
+      "eval_steps_per_second": 1.996,
+      "step": 3280
+    },
+    {
+      "entropy": 0.11450630058534443,
+      "epoch": 8.209215442092155,
+      "grad_norm": 0.19220037758350372,
+      "learning_rate": 1.9113087927659777e-05,
+      "loss": 0.05981999635696411,
+      "mean_token_accuracy": 0.9785088717937469,
+      "num_tokens": 7709745.0,
+      "step": 3300
+    },
+    {
+      "epoch": 8.209215442092155,
+      "eval_entropy": 0.23030704779680386,
+      "eval_loss": 0.9033364653587341,
+      "eval_mean_token_accuracy": 0.8539317641840425,
+      "eval_num_tokens": 7709745.0,
+      "eval_runtime": 86.5804,
+      "eval_samples_per_second": 15.893,
+      "eval_steps_per_second": 1.987,
+      "step": 3300
+    },
+    {
+      "entropy": 0.12258679773658514,
+      "epoch": 8.259028642590286,
+      "grad_norm": 0.1550113558769226,
+      "learning_rate": 1.80999510498665e-05,
+      "loss": 0.06708189249038696,
+      "mean_token_accuracy": 0.9755175545811653,
+      "num_tokens": 7753571.0,
+      "step": 3320
+    },
+    {
+      "epoch": 8.259028642590286,
+      "eval_entropy": 0.22794587762896404,
+      "eval_loss": 0.9068623185157776,
+      "eval_mean_token_accuracy": 0.8544069556302802,
+      "eval_num_tokens": 7753571.0,
+      "eval_runtime": 86.54,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.988,
+      "step": 3320
+    },
+    {
+      "entropy": 0.11094769332557916,
+      "epoch": 8.308841843088418,
+      "grad_norm": 0.17563068866729736,
+      "learning_rate": 1.7111748311196566e-05,
+      "loss": 0.06247819066047668,
+      "mean_token_accuracy": 0.9774838156998158,
+      "num_tokens": 7799310.0,
+      "step": 3340
+    },
+    {
+      "epoch": 8.308841843088418,
+      "eval_entropy": 0.22996614700139956,
+      "eval_loss": 0.9062461853027344,
+      "eval_mean_token_accuracy": 0.8542992434529371,
+      "eval_num_tokens": 7799310.0,
+      "eval_runtime": 86.5131,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3340
+    },
+    {
+      "entropy": 0.11153676519170404,
+      "epoch": 8.35865504358655,
+      "grad_norm": 0.17450673878192902,
+      "learning_rate": 1.61487777401175e-05,
+      "loss": 0.06058149337768555,
+      "mean_token_accuracy": 0.9782821990549564,
+      "num_tokens": 7846447.0,
+      "step": 3360
+    },
+    {
+      "epoch": 8.35865504358655,
+      "eval_entropy": 0.22868362528293631,
+      "eval_loss": 0.9072180390357971,
+      "eval_mean_token_accuracy": 0.8541580455940824,
+      "eval_num_tokens": 7846447.0,
+      "eval_runtime": 86.3954,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 3360
+    },
+    {
+      "entropy": 0.12306302534416318,
+      "epoch": 8.408468244084682,
+      "grad_norm": 0.1664905995130539,
+      "learning_rate": 1.5211329755419379e-05,
+      "loss": 0.06635981798171997,
+      "mean_token_accuracy": 0.9746963500976562,
+      "num_tokens": 7890283.0,
+      "step": 3380
+    },
+    {
+      "epoch": 8.408468244084682,
+      "eval_entropy": 0.22899940029479737,
+      "eval_loss": 0.9086405634880066,
+      "eval_mean_token_accuracy": 0.8544029343959897,
+      "eval_num_tokens": 7890283.0,
+      "eval_runtime": 86.0555,
+      "eval_samples_per_second": 15.99,
+      "eval_steps_per_second": 1.999,
+      "step": 3380
+    },
+    {
+      "entropy": 0.11208283659070731,
+      "epoch": 8.458281444582815,
+      "grad_norm": 0.13771440088748932,
+      "learning_rate": 1.4299687078628192e-05,
+      "loss": 0.06289162039756775,
+      "mean_token_accuracy": 0.9773342818021774,
+      "num_tokens": 7936518.0,
+      "step": 3400
+    },
+    {
+      "epoch": 8.458281444582815,
+      "eval_entropy": 0.23098262182848397,
+      "eval_loss": 0.908290445804596,
+      "eval_mean_token_accuracy": 0.8537809270065885,
+      "eval_num_tokens": 7936518.0,
+      "eval_runtime": 86.7831,
+      "eval_samples_per_second": 15.856,
+      "eval_steps_per_second": 1.982,
+      "step": 3400
+    },
+    {
+      "entropy": 0.12039995044469834,
+      "epoch": 8.508094645080947,
+      "grad_norm": 0.19003549218177795,
+      "learning_rate": 1.3414124648741076e-05,
+      "loss": 0.06479687690734863,
+      "mean_token_accuracy": 0.9768574126064777,
+      "num_tokens": 7981966.0,
+      "step": 3420
+    },
+    {
+      "epoch": 8.508094645080947,
+      "eval_entropy": 0.22869066452217657,
+      "eval_loss": 0.9068775773048401,
+      "eval_mean_token_accuracy": 0.8537808739861776,
+      "eval_num_tokens": 7981966.0,
+      "eval_runtime": 86.6167,
+      "eval_samples_per_second": 15.886,
+      "eval_steps_per_second": 1.986,
+      "step": 3420
+    },
+    {
+      "entropy": 0.11591865480877458,
+      "epoch": 8.557907845579079,
+      "grad_norm": 0.17412032186985016,
+      "learning_rate": 1.2554909539308509e-05,
+      "loss": 0.06449685096740723,
+      "mean_token_accuracy": 0.9773017890751362,
+      "num_tokens": 8028367.0,
+      "step": 3440
+    },
+    {
+      "epoch": 8.557907845579079,
+      "eval_entropy": 0.22861102136761643,
+      "eval_loss": 0.9029546976089478,
+      "eval_mean_token_accuracy": 0.8541917676149413,
+      "eval_num_tokens": 8028367.0,
+      "eval_runtime": 86.5545,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 3440
+    },
+    {
+      "entropy": 0.11211966630071402,
+      "epoch": 8.607721046077211,
+      "grad_norm": 0.17846721410751343,
+      "learning_rate": 1.1722300877888353e-05,
+      "loss": 0.06400871276855469,
+      "mean_token_accuracy": 0.977227509766817,
+      "num_tokens": 8074934.0,
+      "step": 3460
+    },
+    {
+      "epoch": 8.607721046077211,
+      "eval_entropy": 0.2299522036729857,
+      "eval_loss": 0.9003680944442749,
+      "eval_mean_token_accuracy": 0.8541547486255335,
+      "eval_num_tokens": 8074934.0,
+      "eval_runtime": 86.4928,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 3460
+    },
+    {
+      "entropy": 0.11539406068623066,
+      "epoch": 8.657534246575342,
+      "grad_norm": 0.18916265666484833,
+      "learning_rate": 1.0916549767896689e-05,
+      "loss": 0.060999685525894166,
+      "mean_token_accuracy": 0.9764547064900398,
+      "num_tokens": 8122775.0,
+      "step": 3480
+    },
+    {
+      "epoch": 8.657534246575342,
+      "eval_entropy": 0.2297502622701401,
+      "eval_loss": 0.9047583937644958,
+      "eval_mean_token_accuracy": 0.8540163729773012,
+      "eval_num_tokens": 8122775.0,
+      "eval_runtime": 86.5128,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3480
+    },
+    {
+      "entropy": 0.10462163621559739,
+      "epoch": 8.707347447073474,
+      "grad_norm": 0.4441547095775604,
+      "learning_rate": 1.0137899212878408e-05,
+      "loss": 0.0585650622844696,
+      "mean_token_accuracy": 0.9779789052903652,
+      "num_tokens": 8174307.0,
+      "step": 3500
+    },
+    {
+      "epoch": 8.707347447073474,
+      "eval_entropy": 0.2297880021465379,
+      "eval_loss": 0.9056211113929749,
+      "eval_mean_token_accuracy": 0.8541018741075382,
+      "eval_num_tokens": 8174307.0,
+      "eval_runtime": 86.5534,
+      "eval_samples_per_second": 15.898,
+      "eval_steps_per_second": 1.987,
+      "step": 3500
+    },
+    {
+      "entropy": 0.11884278021752834,
+      "epoch": 8.757160647571606,
+      "grad_norm": 0.21446974575519562,
+      "learning_rate": 9.386584043220699e-06,
+      "loss": 0.06243965029716492,
+      "mean_token_accuracy": 0.9768765904009342,
+      "num_tokens": 8222808.0,
+      "step": 3520
+    },
+    {
+      "epoch": 8.757160647571606,
+      "eval_entropy": 0.22878494226308757,
+      "eval_loss": 0.9047290086746216,
+      "eval_mean_token_accuracy": 0.8544826070929683,
+      "eval_num_tokens": 8222808.0,
+      "eval_runtime": 86.6241,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3520
+    },
+    {
+      "entropy": 0.10786062851548195,
+      "epoch": 8.806973848069738,
+      "grad_norm": 0.1993892788887024,
+      "learning_rate": 8.662830845331275e-06,
+      "loss": 0.058893626928329466,
+      "mean_token_accuracy": 0.9780589491128922,
+      "num_tokens": 8273335.0,
+      "step": 3540
+    },
+    {
+      "epoch": 8.806973848069738,
+      "eval_entropy": 0.22842735658551372,
+      "eval_loss": 0.9033588767051697,
+      "eval_mean_token_accuracy": 0.8544253273758777,
+      "eval_num_tokens": 8273335.0,
+      "eval_runtime": 86.5047,
+      "eval_samples_per_second": 15.907,
+      "eval_steps_per_second": 1.988,
+      "step": 3540
+    },
+    {
+      "entropy": 0.12962863883003592,
+      "epoch": 8.85678704856787,
+      "grad_norm": 0.19105808436870575,
+      "learning_rate": 7.96685789330315e-06,
+      "loss": 0.06916859149932861,
+      "mean_token_accuracy": 0.9734630741178989,
+      "num_tokens": 8315235.0,
+      "step": 3560
+    },
+    {
+      "epoch": 8.85678704856787,
+      "eval_entropy": 0.22824140811382337,
+      "eval_loss": 0.9058244228363037,
+      "eval_mean_token_accuracy": 0.8543548379526582,
+      "eval_num_tokens": 8315235.0,
+      "eval_runtime": 86.0706,
+      "eval_samples_per_second": 15.987,
+      "eval_steps_per_second": 1.998,
+      "step": 3560
+    },
+    {
+      "entropy": 0.1322095939423889,
+      "epoch": 8.906600249066003,
+      "grad_norm": 0.1588907092809677,
+      "learning_rate": 7.29887508308606e-06,
+      "loss": 0.0700787365436554,
+      "mean_token_accuracy": 0.9730628825724125,
+      "num_tokens": 8358099.0,
+      "step": 3580
+    },
+    {
+      "epoch": 8.906600249066003,
+      "eval_entropy": 0.22754066853329194,
+      "eval_loss": 0.907920777797699,
+      "eval_mean_token_accuracy": 0.8542819754328839,
+      "eval_num_tokens": 8358099.0,
+      "eval_runtime": 86.6425,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 3580
+    },
+    {
+      "entropy": 0.11120554907247424,
+      "epoch": 8.956413449564135,
+      "grad_norm": 0.15471689403057098,
+      "learning_rate": 6.659083869184757e-06,
+      "loss": 0.06011306643486023,
+      "mean_token_accuracy": 0.9776862308382988,
+      "num_tokens": 8407430.0,
+      "step": 3600
+    },
+    {
+      "epoch": 8.956413449564135,
+      "eval_entropy": 0.22583809070462404,
+      "eval_loss": 0.9105515480041504,
+      "eval_mean_token_accuracy": 0.8541878475699314,
+      "eval_num_tokens": 8407430.0,
+      "eval_runtime": 86.382,
+      "eval_samples_per_second": 15.929,
+      "eval_steps_per_second": 1.991,
+      "step": 3600
+    },
+    {
+      "entropy": 0.11786058268103844,
+      "epoch": 9.004981320049813,
+      "grad_norm": 0.11580188572406769,
+      "learning_rate": 6.047677203903052e-06,
+      "loss": 0.062467324733734134,
+      "mean_token_accuracy": 0.9768382952763484,
+      "num_tokens": 8453175.0,
+      "step": 3620
+    },
+    {
+      "epoch": 9.004981320049813,
+      "eval_entropy": 0.2257705934692261,
+      "eval_loss": 0.9105263352394104,
+      "eval_mean_token_accuracy": 0.8544646265894867,
+      "eval_num_tokens": 8453175.0,
+      "eval_runtime": 86.6206,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 3620
+    },
+    {
+      "entropy": 0.1176839429885149,
+      "epoch": 9.054794520547945,
+      "grad_norm": 0.1505710333585739,
+      "learning_rate": 5.4648394791520385e-06,
+      "loss": 0.06264145970344544,
+      "mean_token_accuracy": 0.9776909500360489,
+      "num_tokens": 8497151.0,
+      "step": 3640
+    },
+    {
+      "epoch": 9.054794520547945,
+      "eval_entropy": 0.22336089827640113,
+      "eval_loss": 0.9186719059944153,
+      "eval_mean_token_accuracy": 0.8543131739594215,
+      "eval_num_tokens": 8497151.0,
+      "eval_runtime": 86.5077,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3640
+    },
+    {
+      "entropy": 0.1169667155481875,
+      "epoch": 9.104607721046078,
+      "grad_norm": 0.13833528757095337,
+      "learning_rate": 4.910746470839993e-06,
+      "loss": 0.059957253932952884,
+      "mean_token_accuracy": 0.9770020335912705,
+      "num_tokens": 8542996.0,
+      "step": 3660
+    },
+    {
+      "epoch": 9.104607721046078,
+      "eval_entropy": 0.2224627248942852,
+      "eval_loss": 0.926105260848999,
+      "eval_mean_token_accuracy": 0.8542806360610696,
+      "eval_num_tokens": 8542996.0,
+      "eval_runtime": 86.6778,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3660
+    },
+    {
+      "entropy": 0.10592180709354579,
+      "epoch": 9.15442092154421,
+      "grad_norm": 0.15626424551010132,
+      "learning_rate": 4.385565285860771e-06,
+      "loss": 0.05797492861747742,
+      "mean_token_accuracy": 0.9791826255619526,
+      "num_tokens": 8591625.0,
+      "step": 3680
+    },
+    {
+      "epoch": 9.15442092154421,
+      "eval_entropy": 0.22266393605359766,
+      "eval_loss": 0.9309298396110535,
+      "eval_mean_token_accuracy": 0.854311544534772,
+      "eval_num_tokens": 8591625.0,
+      "eval_runtime": 86.7402,
+      "eval_samples_per_second": 15.863,
+      "eval_steps_per_second": 1.983,
+      "step": 3680
+    },
+    {
+      "entropy": 0.1101202143356204,
+      "epoch": 9.204234122042342,
+      "grad_norm": 0.16745513677597046,
+      "learning_rate": 3.8894543116967004e-06,
+      "loss": 0.05827975273132324,
+      "mean_token_accuracy": 0.9780424386262894,
+      "num_tokens": 8639845.0,
+      "step": 3700
+    },
+    {
+      "epoch": 9.204234122042342,
+      "eval_entropy": 0.22291506272415781,
+      "eval_loss": 0.9343715906143188,
+      "eval_mean_token_accuracy": 0.8542290634887163,
+      "eval_num_tokens": 8639845.0,
+      "eval_runtime": 86.4609,
+      "eval_samples_per_second": 15.915,
+      "eval_steps_per_second": 1.989,
+      "step": 3700
+    },
+    {
+      "entropy": 0.11024559773504734,
+      "epoch": 9.254047322540472,
+      "grad_norm": 0.19114018976688385,
+      "learning_rate": 3.422563168650925e-06,
+      "loss": 0.059362989664077756,
+      "mean_token_accuracy": 0.9786941170692444,
+      "num_tokens": 8687194.0,
+      "step": 3720
+    },
+    {
+      "epoch": 9.254047322540472,
+      "eval_entropy": 0.22341064467679622,
+      "eval_loss": 0.9363839626312256,
+      "eval_mean_token_accuracy": 0.8541977516440458,
+      "eval_num_tokens": 8687194.0,
+      "eval_runtime": 86.2678,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 3720
+    },
+    {
+      "entropy": 0.12186863766983151,
+      "epoch": 9.303860523038605,
+      "grad_norm": 0.12053123861551285,
+      "learning_rate": 2.985032664724181e-06,
+      "loss": 0.0631858766078949,
+      "mean_token_accuracy": 0.9755026556551456,
+      "num_tokens": 8731721.0,
+      "step": 3740
+    },
+    {
+      "epoch": 9.303860523038605,
+      "eval_entropy": 0.222989774807248,
+      "eval_loss": 0.9386691451072693,
+      "eval_mean_token_accuracy": 0.8542838827815167,
+      "eval_num_tokens": 8731721.0,
+      "eval_runtime": 86.5086,
+      "eval_samples_per_second": 15.906,
+      "eval_steps_per_second": 1.988,
+      "step": 3740
+    },
+    {
+      "entropy": 0.11424487661570311,
+      "epoch": 9.353673723536737,
+      "grad_norm": 0.15845441818237305,
+      "learning_rate": 2.5769947531487898e-06,
+      "loss": 0.058484512567520144,
+      "mean_token_accuracy": 0.9783759318292141,
+      "num_tokens": 8778614.0,
+      "step": 3760
+    },
+    {
+      "epoch": 9.353673723536737,
+      "eval_entropy": 0.22231103705112323,
+      "eval_loss": 0.9413615465164185,
+      "eval_mean_token_accuracy": 0.8542899021575617,
+      "eval_num_tokens": 8778614.0,
+      "eval_runtime": 86.6015,
+      "eval_samples_per_second": 15.889,
+      "eval_steps_per_second": 1.986,
+      "step": 3760
+    },
+    {
+      "entropy": 0.10592789361253381,
+      "epoch": 9.403486924034869,
+      "grad_norm": 0.17964792251586914,
+      "learning_rate": 2.198572492593547e-06,
+      "loss": 0.05741395354270935,
+      "mean_token_accuracy": 0.9798030756413937,
+      "num_tokens": 8826551.0,
+      "step": 3780
+    },
+    {
+      "epoch": 9.403486924034869,
+      "eval_entropy": 0.22196386511935745,
+      "eval_loss": 0.9433969855308533,
+      "eval_mean_token_accuracy": 0.854294165968895,
+      "eval_num_tokens": 8826551.0,
+      "eval_runtime": 86.2332,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 3780
+    },
+    {
+      "entropy": 0.10031416746787727,
+      "epoch": 9.453300124533001,
+      "grad_norm": 0.1429711878299713,
+      "learning_rate": 1.8498800100508529e-06,
+      "loss": 0.05553470253944397,
+      "mean_token_accuracy": 0.9798915058374404,
+      "num_tokens": 8876556.0,
+      "step": 3800
+    },
+    {
+      "epoch": 9.453300124533001,
+      "eval_entropy": 0.22204235256757848,
+      "eval_loss": 0.9440023899078369,
+      "eval_mean_token_accuracy": 0.8542058128257131,
+      "eval_num_tokens": 8876556.0,
+      "eval_runtime": 86.3026,
+      "eval_samples_per_second": 15.944,
+      "eval_steps_per_second": 1.993,
+      "step": 3800
+    },
+    {
+      "entropy": 0.11637815865688025,
+      "epoch": 9.503113325031133,
+      "grad_norm": 0.2096358835697174,
+      "learning_rate": 1.5310224664174822e-06,
+      "loss": 0.06322363018989563,
+      "mean_token_accuracy": 0.9764002375304699,
+      "num_tokens": 8920286.0,
+      "step": 3820
+    },
+    {
+      "epoch": 9.503113325031133,
+      "eval_entropy": 0.2220806197025055,
+      "eval_loss": 0.9451368451118469,
+      "eval_mean_token_accuracy": 0.8541916636533515,
+      "eval_num_tokens": 8920286.0,
+      "eval_runtime": 85.7157,
+      "eval_samples_per_second": 16.053,
+      "eval_steps_per_second": 2.007,
+      "step": 3820
+    },
+    {
+      "entropy": 0.11271723550744354,
+      "epoch": 9.552926525529266,
+      "grad_norm": 0.1656704694032669,
+      "learning_rate": 1.242096024779613e-06,
+      "loss": 0.06295235753059387,
+      "mean_token_accuracy": 0.9773425675928593,
+      "num_tokens": 8964464.0,
+      "step": 3840
+    },
+    {
+      "epoch": 9.552926525529266,
+      "eval_entropy": 0.22215987785264504,
+      "eval_loss": 0.9458719491958618,
+      "eval_mean_token_accuracy": 0.8541110656982245,
+      "eval_num_tokens": 8964464.0,
+      "eval_runtime": 86.5137,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3840
+    },
+    {
+      "entropy": 0.10917426198720932,
+      "epoch": 9.602739726027398,
+      "grad_norm": 0.13864120841026306,
+      "learning_rate": 9.831878214112889e-07,
+      "loss": 0.05977686643600464,
+      "mean_token_accuracy": 0.9781549699604511,
+      "num_tokens": 9010414.0,
+      "step": 3860
+    },
+    {
+      "epoch": 9.602739726027398,
+      "eval_entropy": 0.2222186531438384,
+      "eval_loss": 0.9462024569511414,
+      "eval_mean_token_accuracy": 0.854135605831479,
+      "eval_num_tokens": 9010414.0,
+      "eval_runtime": 86.8131,
+      "eval_samples_per_second": 15.85,
+      "eval_steps_per_second": 1.981,
+      "step": 3860
+    },
+    {
+      "entropy": 0.11271071215160192,
+      "epoch": 9.65255292652553,
+      "grad_norm": 0.11838250607252121,
+      "learning_rate": 7.543759394953306e-07,
+      "loss": 0.05857791900634766,
+      "mean_token_accuracy": 0.9779338933527469,
+      "num_tokens": 9058053.0,
+      "step": 3880
+    },
+    {
+      "epoch": 9.65255292652553,
+      "eval_entropy": 0.22224785959304766,
+      "eval_loss": 0.9465001225471497,
+      "eval_mean_token_accuracy": 0.8541611557783082,
+      "eval_num_tokens": 9058053.0,
+      "eval_runtime": 86.5138,
+      "eval_samples_per_second": 15.905,
+      "eval_steps_per_second": 1.988,
+      "step": 3880
+    },
+    {
+      "entropy": 0.11316167968325316,
+      "epoch": 9.70236612702366,
+      "grad_norm": 0.17645999789237976,
+      "learning_rate": 5.557293855744779e-07,
+      "loss": 0.0591269850730896,
+      "mean_token_accuracy": 0.9781935609877109,
+      "num_tokens": 9105574.0,
+      "step": 3900
+    },
+    {
+      "epoch": 9.70236612702366,
+      "eval_entropy": 0.22210437345296838,
+      "eval_loss": 0.9471818804740906,
+      "eval_mean_token_accuracy": 0.8540260292762933,
+      "eval_num_tokens": 9105574.0,
+      "eval_runtime": 86.584,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.987,
+      "step": 3900
+    },
+    {
+      "entropy": 0.106583485705778,
+      "epoch": 9.752179327521793,
+      "grad_norm": 0.15395089983940125,
+      "learning_rate": 3.873080687399946e-07,
+      "loss": 0.05827186107635498,
+      "mean_token_accuracy": 0.9790284238755703,
+      "num_tokens": 9152793.0,
+      "step": 3920
+    },
+    {
+      "epoch": 9.752179327521793,
+      "eval_entropy": 0.22207596369607505,
+      "eval_loss": 0.9475951194763184,
+      "eval_mean_token_accuracy": 0.8541645428469015,
+      "eval_num_tokens": 9152793.0,
+      "eval_runtime": 86.5265,
+      "eval_samples_per_second": 15.903,
+      "eval_steps_per_second": 1.988,
+      "step": 3920
+    },
+    {
+      "entropy": 0.11190549619495868,
+      "epoch": 9.801992528019925,
+      "grad_norm": 0.2091812938451767,
+      "learning_rate": 2.491627825638833e-07,
+      "loss": 0.060925132036209105,
+      "mean_token_accuracy": 0.9769984439015389,
+      "num_tokens": 9199382.0,
+      "step": 3940
+    },
+    {
+      "epoch": 9.801992528019925,
+      "eval_entropy": 0.221953320468581,
+      "eval_loss": 0.9476391077041626,
+      "eval_mean_token_accuracy": 0.854084026328353,
+      "eval_num_tokens": 9199382.0,
+      "eval_runtime": 86.0341,
+      "eval_samples_per_second": 15.994,
+      "eval_steps_per_second": 1.999,
+      "step": 3940
+    },
+    {
+      "entropy": 0.10218957955949008,
+      "epoch": 9.851805728518057,
+      "grad_norm": 0.12602899968624115,
+      "learning_rate": 1.4133518978026882e-07,
+      "loss": 0.053277283906936646,
+      "mean_token_accuracy": 0.9792918466031552,
+      "num_tokens": 9251912.0,
+      "step": 3960
+    },
+    {
+      "epoch": 9.851805728518057,
+      "eval_entropy": 0.22203073113463645,
+      "eval_loss": 0.9477459192276001,
+      "eval_mean_token_accuracy": 0.8541027304044989,
+      "eval_num_tokens": 9251912.0,
+      "eval_runtime": 86.6774,
+      "eval_samples_per_second": 15.875,
+      "eval_steps_per_second": 1.984,
+      "step": 3960
+    },
+    {
+      "entropy": 0.11482724891975522,
+      "epoch": 9.90161892901619,
+      "grad_norm": 0.1705685257911682,
+      "learning_rate": 6.38578097204261e-08,
+      "loss": 0.061809581518173215,
+      "mean_token_accuracy": 0.9767013140022754,
+      "num_tokens": 9296482.0,
+      "step": 3980
+    },
+    {
+      "epoch": 9.90161892901619,
+      "eval_entropy": 0.22197876593401267,
+      "eval_loss": 0.947809636592865,
+      "eval_mean_token_accuracy": 0.8540589556444523,
+      "eval_num_tokens": 9296482.0,
+      "eval_runtime": 86.9626,
+      "eval_samples_per_second": 15.823,
+      "eval_steps_per_second": 1.978,
+      "step": 3980
+    },
+    {
+      "entropy": 0.1206501518841833,
+      "epoch": 9.951432129514322,
+      "grad_norm": 0.17280042171478271,
+      "learning_rate": 1.675400850544711e-08,
+      "loss": 0.06324169635772706,
+      "mean_token_accuracy": 0.9753918401896954,
+      "num_tokens": 9341118.0,
+      "step": 4000
+    },
+    {
+      "epoch": 9.951432129514322,
+      "eval_entropy": 0.22201869714745257,
+      "eval_loss": 0.9477588534355164,
+      "eval_mean_token_accuracy": 0.8541659501402877,
+      "eval_num_tokens": 9341118.0,
+      "eval_runtime": 86.6633,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 4000
+    },
+    {
+      "entropy": 0.12383906352214324,
+      "epoch": 10.0,
+      "grad_norm": 0.306317538022995,
+      "learning_rate": 3.799199930972323e-11,
+      "loss": 0.058901286125183104,
+      "mean_token_accuracy": 0.9756807119418414,
+      "num_tokens": 9386580.0,
+      "step": 4020
+    },
+    {
+      "epoch": 10.0,
+      "eval_entropy": 0.22198900363819543,
+      "eval_loss": 0.9476598501205444,
+      "eval_mean_token_accuracy": 0.8541736436444659,
+      "eval_num_tokens": 9386580.0,
+      "eval_runtime": 86.5843,
+      "eval_samples_per_second": 15.892,
+      "eval_steps_per_second": 1.987,
+      "step": 4020
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.964083844973138e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..de447ba6f446520bcf43431893d5eee022e98f39
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-420/trainer_state.json
@@ -0,0 +1,475 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0448318804483188,
+  "eval_steps": 20,
+  "global_step": 420,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.155416952633139e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..35d6869023775ee317b0007965e377dbf888b4a5
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-440/trainer_state.json
@@ -0,0 +1,496 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0946450809464507,
+  "eval_steps": 20,
+  "global_step": 440,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.352373181726925e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7eafefe11c9eb1ce15be7f1da15604f08b7ecb90
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-460/trainer_state.json
@@ -0,0 +1,517 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.1444582814445827,
+  "eval_steps": 20,
+  "global_step": 460,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.572855012445594e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..731d27270f263e6b46db130f9a6bfe89aeecb020
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-480/trainer_state.json
@@ -0,0 +1,538 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.1942714819427147,
+  "eval_steps": 20,
+  "global_step": 480,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.77945126296064e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6389949bb47e969fcc62784a76023708ccbaac58
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-500/trainer_state.json
@@ -0,0 +1,559 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.244084682440847,
+  "eval_steps": 20,
+  "global_step": 500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.964191412767334e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7431c54cca80d7bf5d7ad819af6c7d4baf436fec
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-520/trainer_state.json
@@ -0,0 +1,580 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.293897882938979,
+  "eval_steps": 20,
+  "global_step": 520,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.160321867214234e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..adbedab2f6095a227798d5c429279993de6e659b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-540/trainer_state.json
@@ -0,0 +1,601 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.3437110834371109,
+  "eval_steps": 20,
+  "global_step": 540,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.344020822900941e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b06e49f91017a1bc8c9e54698ec92ebedc51faa
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-560/trainer_state.json
@@ -0,0 +1,622 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.3935242839352429,
+  "eval_steps": 20,
+  "global_step": 560,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.543921118127104e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c0eea164a5d0f00967f61dd955c0f7059f14d13
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-580/trainer_state.json
@@ -0,0 +1,643 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.4433374844333748,
+  "eval_steps": 20,
+  "global_step": 580,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.739225797927117e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea6de11fd9cb1d92d79d85d67a11b3a90f245ca3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-60/trainer_state.json
@@ -0,0 +1,97 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.149439601494396,
+  "eval_steps": 20,
+  "global_step": 60,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5653773829644288.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7850ce4e7bf17a6653a900d5de7b1e97c33e8b3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-600/trainer_state.json
@@ -0,0 +1,664 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.4931506849315068,
+  "eval_steps": 20,
+  "global_step": 600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.93531137331712e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..dcd6566a7e4d0ed9ad244e0578ef8080382bc024
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-620/trainer_state.json
@@ -0,0 +1,685 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.5429638854296388,
+  "eval_steps": 20,
+  "global_step": 620,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.147230279980032e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..b41e77b995750501cbf50620ab406e2f24d9cd5a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-640/trainer_state.json
@@ -0,0 +1,706 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.592777085927771,
+  "eval_steps": 20,
+  "global_step": 640,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.338594578584576e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..06e962b8b27da5880c2f536efc26c248639459b8
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-660/trainer_state.json
@@ -0,0 +1,727 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.6425902864259028,
+  "eval_steps": 20,
+  "global_step": 660,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.516450281063424e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..11a5aab197c8310a23a8c88a8487a9bc0bb007c1
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-680/trainer_state.json
@@ -0,0 +1,748 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.692403486924035,
+  "eval_steps": 20,
+  "global_step": 680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.728503824897024e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1b0d804d310f67c95b148fec716180c030de63b
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-700/trainer_state.json
@@ -0,0 +1,769 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.7422166874221667,
+  "eval_steps": 20,
+  "global_step": 700,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 6.93499236567552e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..db673f63554676398607aed8940e354eeceab60e
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-720/trainer_state.json
@@ -0,0 +1,790 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.792029887920299,
+  "eval_steps": 20,
+  "global_step": 720,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.120576241751859e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..681681c890db61702c5048dcc8bc598e7a4e58ac
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-740/trainer_state.json
@@ -0,0 +1,811 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.841843088418431,
+  "eval_steps": 20,
+  "global_step": 740,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.32430920843694e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a26c7bace27c5d89d459c69597b9ba01b7ab9882
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-760/trainer_state.json
@@ -0,0 +1,832 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.891656288916563,
+  "eval_steps": 20,
+  "global_step": 760,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.507783768839168e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..470a593a906de5298077be1b2ea770afefd21749
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-780/trainer_state.json
@@ -0,0 +1,853 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9414694894146949,
+  "eval_steps": 20,
+  "global_step": 780,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.706427450472243e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..eacb6ea661208cb0cf708122b8a2577f65627538
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-80/trainer_state.json
@@ -0,0 +1,118 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.199252801992528,
+  "eval_steps": 20,
+  "global_step": 80,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7798902991159296.0,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..021ad51045af2d13b89172cbe7a3061cd0bbfe7c
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-800/trainer_state.json
@@ -0,0 +1,874 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9912826899128269,
+  "eval_steps": 20,
+  "global_step": 800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.897684039340237e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..6bca6d503fd288ba87725cbdf02385d68b4810fe
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-820/trainer_state.json
@@ -0,0 +1,895 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0398505603985058,
+  "eval_steps": 20,
+  "global_step": 820,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.08591129186775e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ceb1ee53a4519906d649ea02aac61d5bd09e02c6
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-840/trainer_state.json
@@ -0,0 +1,916 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.0896637608966375,
+  "eval_steps": 20,
+  "global_step": 840,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.296977496449638e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c1f15076e17560f79837bf6f1e99764620919e3
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-860/trainer_state.json
@@ -0,0 +1,937 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.1394769613947697,
+  "eval_steps": 20,
+  "global_step": 860,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.478540209028096e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..0606a8f7c430492a184962b786b1e29f0b1c524a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-880/trainer_state.json
@@ -0,0 +1,958 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.1892901618929015,
+  "eval_steps": 20,
+  "global_step": 880,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.682066732051456e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f24191a0827c29a138d60001fc4b19cee495dff
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-900/trainer_state.json
@@ -0,0 +1,979 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.2391033623910337,
+  "eval_steps": 20,
+  "global_step": 900,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 8.875351854291149e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..ead149317a3e844b794f81b82ba71e463b164e3a
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-920/trainer_state.json
@@ -0,0 +1,1000 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.2889165628891655,
+  "eval_steps": 20,
+  "global_step": 920,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.073672406714573e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d717a898943078691b1e34460d294d531ca3ed7
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-940/trainer_state.json
@@ -0,0 +1,1021 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.3387297633872977,
+  "eval_steps": 20,
+  "global_step": 940,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.248826389968282e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..2dba324418ead5931a17d6a764182c3e1b663683
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-960/trainer_state.json
@@ -0,0 +1,1042 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.3885429638854294,
+  "eval_steps": 20,
+  "global_step": 960,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.454911019234714e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/README.md b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..41e6c854e77830e9ea767c8c35f8c82a65c1ba35
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/README.md
@@ -0,0 +1,209 @@
+---
+base_model: Qwen/Qwen3.5-4B-Base
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3.5-4B-Base
+- lora
+- sft
+- transformers
+- trl
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.18.1
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/adapter_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb037202d6e02304d118bbaf965155df8df30c18
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/adapter_config.json
@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3.5-4B-Base",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.0005183818805460705,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "up_proj",
+    "q_proj",
+    "o_proj",
+    "v_proj",
+    "k_proj",
+    "gate_proj",
+    "down_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/chat_template.jinja b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/chat_template.jinja
new file mode 100644
index 0000000000000000000000000000000000000000..a585dec894e63da457d9440ec6aa7caa16d20860
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/chat_template.jinja
@@ -0,0 +1,154 @@
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- macro render_content(content, do_vision_count, is_system_content=false) %}
+    {%- if content is string %}
+        {{- content }}
+    {%- elif content is iterable and content is not mapping %}
+        {%- for item in content %}
+            {%- if 'image' in item or 'image_url' in item or item.type == 'image' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain images.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Picture ' ~ image_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|image_pad|><|vision_end|>' }}
+            {%- elif 'video' in item or item.type == 'video' %}
+                {%- if is_system_content %}
+                    {{- raise_exception('System message cannot contain videos.') }}
+                {%- endif %}
+                {%- if do_vision_count %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                {%- endif %}
+                {%- if add_vision_id %}
+                    {{- 'Video ' ~ video_count.value ~ ': ' }}
+                {%- endif %}
+                {{- '<|vision_start|><|video_pad|><|vision_end|>' }}
+            {%- elif 'text' in item %}
+                {{- item.text }}
+            {%- else %}
+                {{- raise_exception('Unexpected item type in content.') }}
+            {%- endif %}
+        {%- endfor %}
+    {%- elif content is none or content is undefined %}
+        {{- '' }}
+    {%- else %}
+        {{- raise_exception('Unexpected content type.') }}
+    {%- endif %}
+{%- endmacro %}
+{%- if not messages %}
+    {{- raise_exception('No messages provided.') }}
+{%- endif %}
+{%- if tools and tools is iterable and tools is not mapping %}
+    {{- '<|im_start|>system\n' }}
+    {{- "# Tools\n\nYou have access to the following functions:\n\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>" }}
+    {{- '\n\nIf you choose to call a function ONLY reply in the following format with NO suffix:\n\n<tool_call>\n<function=example_function_name>\n<parameter=example_parameter_1>\nvalue_1\n</parameter>\n<parameter=example_parameter_2>\nThis is the value for the second parameter\nthat can span\nmultiple lines\n</parameter>\n</function>\n</tool_call>\n\n<IMPORTANT>\nReminder:\n- Function calls MUST follow the specified format: an inner <function=...></function> block must be nested within <tool_call></tool_call> XML tags\n- Required parameters MUST be specified\n- You may provide optional reasoning for your function call in natural language BEFORE the function call, but NOT after\n- If there is no function call available, answer the question like normal with your current knowledge and do not tell the user about function calls\n</IMPORTANT>' }}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {%- if content %}
+            {{- '\n\n' + content }}
+        {%- endif %}
+    {%- endif %}
+    {{- '<|im_end|>\n' }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {%- set content = render_content(messages[0].content, false, true)|trim %}
+        {{- '<|im_start|>system\n' + content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}
+{%- for message in messages[::-1] %}
+    {%- set index = (messages|length - 1) - loop.index0 %}
+    {%- if ns.multi_step_tool and message.role == "user" %}
+        {%- set content = render_content(message.content, false)|trim %}
+        {%- if not(content.startswith('<tool_response>') and content.endswith('</tool_response>')) %}
+            {%- set ns.multi_step_tool = false %}
+            {%- set ns.last_query_index = index %}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if ns.multi_step_tool %}
+    {{- raise_exception('No user query found in messages.') }}
+{%- endif %}
+{%- for message in messages %}
+    {%- set content = render_content(message.content, true)|trim %}
+    {%- if message.role == "system" %}
+        {%- if not loop.first %}
+            {{- raise_exception('System message must be at the beginning.') }}
+        {%- endif %}
+    {%- elif message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {%- set reasoning_content = '' %}
+        {%- if message.reasoning_content is string %}
+            {%- set reasoning_content = message.reasoning_content %}
+        {%- else %}
+            {%- if '</think>' in content %}
+                {%- set reasoning_content = content.split('</think>')[0].rstrip('\n').split('<think>')[-1].lstrip('\n') %}
+                {%- set content = content.split('</think>')[-1].lstrip('\n') %}
+            {%- endif %}
+        {%- endif %}
+        {%- set reasoning_content = reasoning_content|trim %}
+        {%- if loop.index0 > ns.last_query_index %}
+            {{- '<|im_start|>' + message.role + '\n<think>\n' + reasoning_content + '\n</think>\n\n' + content }}
+        {%- else %}
+            {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- endif %}
+        {%- if message.tool_calls and message.tool_calls is iterable and message.tool_calls is not mapping %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if tool_call.function is defined %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {%- if loop.first %}
+                    {%- if content|trim %}
+                        {{- '\n\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- else %}
+                        {{- '<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                    {%- endif %}
+                {%- else %}
+                    {{- '\n<tool_call>\n<function=' + tool_call.name + '>\n' }}
+                {%- endif %}
+                {%- if tool_call.arguments is defined %}
+                    {%- for args_name, args_value in tool_call.arguments|items %}
+                        {{- '<parameter=' + args_name + '>\n' }}
+                        {%- set args_value = args_value | tojson | safe if args_value is mapping or (args_value is sequence and args_value is not string) else args_value | string %}
+                        {{- args_value }}
+                        {{- '\n</parameter>\n' }}
+                    {%- endfor %}
+                {%- endif %}
+                {{- '</function>\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.previtem and loop.previtem.role != "tool" %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if not loop.last and loop.nextitem.role != "tool" %}
+            {{- '<|im_end|>\n' }}
+        {%- elif loop.last %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- else %}
+        {{- raise_exception('Unexpected message role.') }}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+    {%- if enable_thinking is defined and enable_thinking is false %}
+        {{- '<think>\n\n</think>\n\n' }}
+    {%- else %}
+        {{- '<think>\n' }}
+    {%- endif %}
+{%- endif %}
\ No newline at end of file
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/tokenizer_config.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a37b2a6fd3ab3317cd7bac72855be1a843b2bb
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/tokenizer_config.json
@@ -0,0 +1,31 @@
+{
+  "add_prefix_space": false,
+  "audio_bos_token": "<|audio_start|>",
+  "audio_eos_token": "<|audio_end|>",
+  "audio_token": "<|audio_pad|>",
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|endoftext|>",
+  "errors": "replace",
+  "image_token": "<|image_pad|>",
+  "is_local": false,
+  "model_max_length": 262144,
+  "model_specific_special_tokens": {
+    "audio_bos_token": "<|audio_start|>",
+    "audio_eos_token": "<|audio_end|>",
+    "audio_token": "<|audio_pad|>",
+    "image_token": "<|image_pad|>",
+    "video_token": "<|video_pad|>",
+    "vision_bos_token": "<|vision_start|>",
+    "vision_eos_token": "<|vision_end|>"
+  },
+  "pad_token": "<|endoftext|>",
+  "pretokenize_regex": "(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?[\\p{L}\\p{M}]+|\\p{N}| ?[^\\s\\p{L}\\p{M}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+  "split_special_tokens": false,
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null,
+  "video_token": "<|video_pad|>",
+  "vision_bos_token": "<|vision_start|>",
+  "vision_eos_token": "<|vision_end|>"
+}
diff --git a/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/trainer_state.json b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..29f99b01ea4f84c14abc3eb1a74a7ef50d7c0ca1
--- /dev/null
+++ b/overgeneralisation_original_Swedish/Qwen3.5-4B-Base_overgeneralisation_splits_original_features_train_overgeneralisation_splits_original_features_test1/checkpoint-980/trainer_state.json
@@ -0,0 +1,1063 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.4383561643835616,
+  "eval_steps": 20,
+  "global_step": 980,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "entropy": 1.9784346982836722,
+      "epoch": 0.049813200498132,
+      "grad_norm": 3.0229668617248535,
+      "learning_rate": 9.526142962415369e-06,
+      "loss": 1.7360023498535155,
+      "mean_token_accuracy": 0.6449888605624438,
+      "num_tokens": 46794.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.049813200498132,
+      "eval_entropy": 1.41506897571475,
+      "eval_loss": 1.1876318454742432,
+      "eval_mean_token_accuracy": 0.734131895525511,
+      "eval_num_tokens": 46794.0,
+      "eval_runtime": 87.8071,
+      "eval_samples_per_second": 15.671,
+      "eval_steps_per_second": 1.959,
+      "step": 20
+    },
+    {
+      "entropy": 1.049924298375845,
+      "epoch": 0.099626400996264,
+      "grad_norm": 1.5795097351074219,
+      "learning_rate": 1.9553661870221022e-05,
+      "loss": 0.8944448471069336,
+      "mean_token_accuracy": 0.7748479396104813,
+      "num_tokens": 90754.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.099626400996264,
+      "eval_entropy": 0.7996658658565476,
+      "eval_loss": 0.7202735543251038,
+      "eval_mean_token_accuracy": 0.8070558306089667,
+      "eval_num_tokens": 90754.0,
+      "eval_runtime": 86.9199,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 1.979,
+      "step": 40
+    },
+    {
+      "entropy": 0.7734908878803253,
+      "epoch": 0.149439601494396,
+      "grad_norm": 1.3136248588562012,
+      "learning_rate": 2.9581180778026673e-05,
+      "loss": 0.6780608654022217,
+      "mean_token_accuracy": 0.8168170280754566,
+      "num_tokens": 137472.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.149439601494396,
+      "eval_entropy": 0.7119324009778888,
+      "eval_loss": 0.6554311513900757,
+      "eval_mean_token_accuracy": 0.8215604798738346,
+      "eval_num_tokens": 137472.0,
+      "eval_runtime": 86.8692,
+      "eval_samples_per_second": 15.84,
+      "eval_steps_per_second": 1.98,
+      "step": 60
+    },
+    {
+      "entropy": 0.7071127541363239,
+      "epoch": 0.199252801992528,
+      "grad_norm": 1.387060284614563,
+      "learning_rate": 3.960869968583232e-05,
+      "loss": 0.6382100582122803,
+      "mean_token_accuracy": 0.8229366384446621,
+      "num_tokens": 187408.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.199252801992528,
+      "eval_entropy": 0.6883931482254073,
+      "eval_loss": 0.625065803527832,
+      "eval_mean_token_accuracy": 0.828940509710201,
+      "eval_num_tokens": 187408.0,
+      "eval_runtime": 86.662,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 80
+    },
+    {
+      "entropy": 0.6800824083387852,
+      "epoch": 0.24906600249066002,
+      "grad_norm": 0.9892916679382324,
+      "learning_rate": 4.963621859363797e-05,
+      "loss": 0.6011715888977051,
+      "mean_token_accuracy": 0.8323964163661003,
+      "num_tokens": 234197.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.24906600249066002,
+      "eval_entropy": 0.6840810470802839,
+      "eval_loss": 0.6037028431892395,
+      "eval_mean_token_accuracy": 0.8309669033732525,
+      "eval_num_tokens": 234197.0,
+      "eval_runtime": 86.4637,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 100
+    },
+    {
+      "entropy": 0.6776216626167297,
+      "epoch": 0.298879202988792,
+      "grad_norm": 0.8918434977531433,
+      "learning_rate": 5.9663737501443624e-05,
+      "loss": 0.5991742610931396,
+      "mean_token_accuracy": 0.8300838828086853,
+      "num_tokens": 281241.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.298879202988792,
+      "eval_entropy": 0.690427724705186,
+      "eval_loss": 0.5939701795578003,
+      "eval_mean_token_accuracy": 0.8345950186945671,
+      "eval_num_tokens": 281241.0,
+      "eval_runtime": 86.6626,
+      "eval_samples_per_second": 15.878,
+      "eval_steps_per_second": 1.985,
+      "step": 120
+    },
+    {
+      "entropy": 0.6709842771291733,
+      "epoch": 0.34869240348692404,
+      "grad_norm": 0.9135531187057495,
+      "learning_rate": 6.969125640924927e-05,
+      "loss": 0.5914147377014161,
+      "mean_token_accuracy": 0.8314545609056949,
+      "num_tokens": 327393.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.34869240348692404,
+      "eval_entropy": 0.6584504666023476,
+      "eval_loss": 0.5849721431732178,
+      "eval_mean_token_accuracy": 0.8357757236375365,
+      "eval_num_tokens": 327393.0,
+      "eval_runtime": 86.3262,
+      "eval_samples_per_second": 15.94,
+      "eval_steps_per_second": 1.992,
+      "step": 140
+    },
+    {
+      "entropy": 0.6524647936224938,
+      "epoch": 0.398505603985056,
+      "grad_norm": 0.8651587963104248,
+      "learning_rate": 7.971877531705493e-05,
+      "loss": 0.5710843563079834,
+      "mean_token_accuracy": 0.8396127380430698,
+      "num_tokens": 373834.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.398505603985056,
+      "eval_entropy": 0.6283470298661742,
+      "eval_loss": 0.5738973617553711,
+      "eval_mean_token_accuracy": 0.8379981181649274,
+      "eval_num_tokens": 373834.0,
+      "eval_runtime": 86.5619,
+      "eval_samples_per_second": 15.896,
+      "eval_steps_per_second": 1.987,
+      "step": 160
+    },
+    {
+      "entropy": 0.6450445972383022,
+      "epoch": 0.44831880448318806,
+      "grad_norm": 0.8661723732948303,
+      "learning_rate": 8.974629422486058e-05,
+      "loss": 0.5677794933319091,
+      "mean_token_accuracy": 0.8389350369572639,
+      "num_tokens": 422572.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.44831880448318806,
+      "eval_entropy": 0.6142613257086554,
+      "eval_loss": 0.5698265433311462,
+      "eval_mean_token_accuracy": 0.8388577273418737,
+      "eval_num_tokens": 422572.0,
+      "eval_runtime": 86.4443,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 180
+    },
+    {
+      "entropy": 0.6448334597051144,
+      "epoch": 0.49813200498132004,
+      "grad_norm": 0.9662242531776428,
+      "learning_rate": 9.977381313266624e-05,
+      "loss": 0.581433916091919,
+      "mean_token_accuracy": 0.8387043006718159,
+      "num_tokens": 471879.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.49813200498132004,
+      "eval_entropy": 0.6154296522916749,
+      "eval_loss": 0.5660303831100464,
+      "eval_mean_token_accuracy": 0.8412494766850804,
+      "eval_num_tokens": 471879.0,
+      "eval_runtime": 86.3063,
+      "eval_samples_per_second": 15.943,
+      "eval_steps_per_second": 1.993,
+      "step": 200
+    },
+    {
+      "entropy": 0.6376728117465973,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.7618638873100281,
+      "learning_rate": 0.00010980133204047189,
+      "loss": 0.5678351402282715,
+      "mean_token_accuracy": 0.8404546812176704,
+      "num_tokens": 520984.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.547945205479452,
+      "eval_entropy": 0.6181817033956217,
+      "eval_loss": 0.5663750171661377,
+      "eval_mean_token_accuracy": 0.8388350962899452,
+      "eval_num_tokens": 520984.0,
+      "eval_runtime": 86.5904,
+      "eval_samples_per_second": 15.891,
+      "eval_steps_per_second": 1.986,
+      "step": 220
+    },
+    {
+      "entropy": 0.6303176879882812,
+      "epoch": 0.597758405977584,
+      "grad_norm": 0.7571695446968079,
+      "learning_rate": 0.00011982885094827753,
+      "loss": 0.5502778053283691,
+      "mean_token_accuracy": 0.8429657347500324,
+      "num_tokens": 566596.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.597758405977584,
+      "eval_entropy": 0.6252533817707107,
+      "eval_loss": 0.5570284128189087,
+      "eval_mean_token_accuracy": 0.8427327847064927,
+      "eval_num_tokens": 566596.0,
+      "eval_runtime": 86.4157,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 240
+    },
+    {
+      "entropy": 0.6202544964849949,
+      "epoch": 0.6475716064757161,
+      "grad_norm": 0.6447190642356873,
+      "learning_rate": 0.00012985636985608318,
+      "loss": 0.5485352993011474,
+      "mean_token_accuracy": 0.844165726006031,
+      "num_tokens": 613603.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.6475716064757161,
+      "eval_entropy": 0.6441633552312851,
+      "eval_loss": 0.5606644153594971,
+      "eval_mean_token_accuracy": 0.842403513054515,
+      "eval_num_tokens": 613603.0,
+      "eval_runtime": 86.6343,
+      "eval_samples_per_second": 15.883,
+      "eval_steps_per_second": 1.985,
+      "step": 260
+    },
+    {
+      "entropy": 0.6306711677461863,
+      "epoch": 0.6973848069738481,
+      "grad_norm": 0.7869907021522522,
+      "learning_rate": 0.00013988388876388883,
+      "loss": 0.5579307556152344,
+      "mean_token_accuracy": 0.841247134655714,
+      "num_tokens": 658565.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.6973848069738481,
+      "eval_entropy": 0.6263934678809587,
+      "eval_loss": 0.5559113025665283,
+      "eval_mean_token_accuracy": 0.8427334743183713,
+      "eval_num_tokens": 658565.0,
+      "eval_runtime": 86.6403,
+      "eval_samples_per_second": 15.882,
+      "eval_steps_per_second": 1.985,
+      "step": 280
+    },
+    {
+      "entropy": 0.6385872110724449,
+      "epoch": 0.7471980074719801,
+      "grad_norm": 0.6679229736328125,
+      "learning_rate": 0.0001499114076716945,
+      "loss": 0.5667279720306396,
+      "mean_token_accuracy": 0.8389136254787445,
+      "num_tokens": 705680.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.7471980074719801,
+      "eval_entropy": 0.6141417321077612,
+      "eval_loss": 0.5570600628852844,
+      "eval_mean_token_accuracy": 0.8437647996253745,
+      "eval_num_tokens": 705680.0,
+      "eval_runtime": 86.7588,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 300
+    },
+    {
+      "entropy": 0.6199494235217571,
+      "epoch": 0.797011207970112,
+      "grad_norm": 0.7924400568008423,
+      "learning_rate": 0.00015993892657950015,
+      "loss": 0.5529299736022949,
+      "mean_token_accuracy": 0.8426973208785057,
+      "num_tokens": 752616.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.797011207970112,
+      "eval_entropy": 0.6133768925833147,
+      "eval_loss": 0.556602418422699,
+      "eval_mean_token_accuracy": 0.8432947965555413,
+      "eval_num_tokens": 752616.0,
+      "eval_runtime": 86.492,
+      "eval_samples_per_second": 15.909,
+      "eval_steps_per_second": 1.989,
+      "step": 320
+    },
+    {
+      "entropy": 0.6203986253589392,
+      "epoch": 0.8468244084682441,
+      "grad_norm": 0.8364354372024536,
+      "learning_rate": 0.00016996644548730578,
+      "loss": 0.5551144123077393,
+      "mean_token_accuracy": 0.8432973213493824,
+      "num_tokens": 797151.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.8468244084682441,
+      "eval_entropy": 0.6017442844634833,
+      "eval_loss": 0.5566568374633789,
+      "eval_mean_token_accuracy": 0.8437666123689607,
+      "eval_num_tokens": 797151.0,
+      "eval_runtime": 86.5552,
+      "eval_samples_per_second": 15.897,
+      "eval_steps_per_second": 1.987,
+      "step": 340
+    },
+    {
+      "entropy": 0.6341533534228802,
+      "epoch": 0.8966376089663761,
+      "grad_norm": 0.7783445715904236,
+      "learning_rate": 0.00017999396439511144,
+      "loss": 0.5669133186340332,
+      "mean_token_accuracy": 0.8379446342587471,
+      "num_tokens": 843585.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.8966376089663761,
+      "eval_entropy": 0.6055107958788095,
+      "eval_loss": 0.5599350333213806,
+      "eval_mean_token_accuracy": 0.8435030894917112,
+      "eval_num_tokens": 843585.0,
+      "eval_runtime": 86.4814,
+      "eval_samples_per_second": 15.911,
+      "eval_steps_per_second": 1.989,
+      "step": 360
+    },
+    {
+      "entropy": 0.6306198488920927,
+      "epoch": 0.9464508094645081,
+      "grad_norm": 0.8449786901473999,
+      "learning_rate": 0.0001900214833029171,
+      "loss": 0.5739435195922852,
+      "mean_token_accuracy": 0.8393832489848136,
+      "num_tokens": 889842.0,
+      "step": 380
+    },
+    {
+      "epoch": 0.9464508094645081,
+      "eval_entropy": 0.6129532439071078,
+      "eval_loss": 0.5566295981407166,
+      "eval_mean_token_accuracy": 0.8430350880290187,
+      "eval_num_tokens": 889842.0,
+      "eval_runtime": 86.4643,
+      "eval_samples_per_second": 15.914,
+      "eval_steps_per_second": 1.989,
+      "step": 380
+    },
+    {
+      "entropy": 0.6203123550862074,
+      "epoch": 0.9962640099626401,
+      "grad_norm": 0.7334314584732056,
+      "learning_rate": 0.00020004900221072276,
+      "loss": 0.5547565937042236,
+      "mean_token_accuracy": 0.8403573960065842,
+      "num_tokens": 935589.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.9962640099626401,
+      "eval_entropy": 0.6275761647279873,
+      "eval_loss": 0.5621116757392883,
+      "eval_mean_token_accuracy": 0.841587379228237,
+      "eval_num_tokens": 935589.0,
+      "eval_runtime": 86.4748,
+      "eval_samples_per_second": 15.912,
+      "eval_steps_per_second": 1.989,
+      "step": 400
+    },
+    {
+      "entropy": 0.5795013002860241,
+      "epoch": 1.0448318804483188,
+      "grad_norm": 0.8858296871185303,
+      "learning_rate": 0.0002015421505577756,
+      "loss": 0.5183939933776855,
+      "mean_token_accuracy": 0.850081592034071,
+      "num_tokens": 980589.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.0448318804483188,
+      "eval_entropy": 0.5583065545489622,
+      "eval_loss": 0.5605642199516296,
+      "eval_mean_token_accuracy": 0.8439708411000496,
+      "eval_num_tokens": 980589.0,
+      "eval_runtime": 86.5422,
+      "eval_samples_per_second": 15.9,
+      "eval_steps_per_second": 1.987,
+      "step": 420
+    },
+    {
+      "entropy": 0.5671238023787737,
+      "epoch": 1.0946450809464507,
+      "grad_norm": 0.6882498264312744,
+      "learning_rate": 0.00020150112347025443,
+      "loss": 0.5077326774597168,
+      "mean_token_accuracy": 0.8489868573844432,
+      "num_tokens": 1027852.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.0946450809464507,
+      "eval_entropy": 0.5868900277933409,
+      "eval_loss": 0.5602695345878601,
+      "eval_mean_token_accuracy": 0.8428842161977014,
+      "eval_num_tokens": 1027852.0,
+      "eval_runtime": 86.623,
+      "eval_samples_per_second": 15.885,
+      "eval_steps_per_second": 1.986,
+      "step": 440
+    },
+    {
+      "entropy": 0.5533561781048775,
+      "epoch": 1.1444582814445827,
+      "grad_norm": 0.7717723250389099,
+      "learning_rate": 0.0002014297192297181,
+      "loss": 0.4954517364501953,
+      "mean_token_accuracy": 0.8529035650193691,
+      "num_tokens": 1077649.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.1444582814445827,
+      "eval_entropy": 0.5600803743961246,
+      "eval_loss": 0.5608077645301819,
+      "eval_mean_token_accuracy": 0.8445036771685578,
+      "eval_num_tokens": 1077649.0,
+      "eval_runtime": 86.1316,
+      "eval_samples_per_second": 15.976,
+      "eval_steps_per_second": 1.997,
+      "step": 460
+    },
+    {
+      "entropy": 0.5692154694348573,
+      "epoch": 1.1942714819427147,
+      "grad_norm": 0.7322827577590942,
+      "learning_rate": 0.0002013279593707117,
+      "loss": 0.505049467086792,
+      "mean_token_accuracy": 0.8551576808094978,
+      "num_tokens": 1124872.0,
+      "step": 480
+    },
+    {
+      "epoch": 1.1942714819427147,
+      "eval_entropy": 0.5732695829383162,
+      "eval_loss": 0.5594323873519897,
+      "eval_mean_token_accuracy": 0.8449713407560836,
+      "eval_num_tokens": 1124872.0,
+      "eval_runtime": 86.2726,
+      "eval_samples_per_second": 15.949,
+      "eval_steps_per_second": 1.994,
+      "step": 480
+    },
+    {
+      "entropy": 0.5817618492990733,
+      "epoch": 1.244084682440847,
+      "grad_norm": 1.1776764392852783,
+      "learning_rate": 0.0002011958745826208,
+      "loss": 0.5137609958648681,
+      "mean_token_accuracy": 0.8521522544324398,
+      "num_tokens": 1168698.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.244084682440847,
+      "eval_entropy": 0.5662581343636957,
+      "eval_loss": 0.5595026016235352,
+      "eval_mean_token_accuracy": 0.8441977164773053,
+      "eval_num_tokens": 1168698.0,
+      "eval_runtime": 86.7261,
+      "eval_samples_per_second": 15.866,
+      "eval_steps_per_second": 1.983,
+      "step": 500
+    },
+    {
+      "entropy": 0.5712925456464291,
+      "epoch": 1.293897882938979,
+      "grad_norm": 0.7960361838340759,
+      "learning_rate": 0.0002010335047004159,
+      "loss": 0.5134767532348633,
+      "mean_token_accuracy": 0.8513577707111836,
+      "num_tokens": 1216679.0,
+      "step": 520
+    },
+    {
+      "epoch": 1.293897882938979,
+      "eval_entropy": 0.5441222797299541,
+      "eval_loss": 0.5535460114479065,
+      "eval_mean_token_accuracy": 0.8450886118550633,
+      "eval_num_tokens": 1216679.0,
+      "eval_runtime": 86.2675,
+      "eval_samples_per_second": 15.95,
+      "eval_steps_per_second": 1.994,
+      "step": 520
+    },
+    {
+      "entropy": 0.5787045754492283,
+      "epoch": 1.3437110834371109,
+      "grad_norm": 0.9205410480499268,
+      "learning_rate": 0.00020084089869263887,
+      "loss": 0.5119701862335205,
+      "mean_token_accuracy": 0.8503516331315041,
+      "num_tokens": 1261365.0,
+      "step": 540
+    },
+    {
+      "epoch": 1.3437110834371109,
+      "eval_entropy": 0.5744457827057949,
+      "eval_loss": 0.5514978766441345,
+      "eval_mean_token_accuracy": 0.845929987901865,
+      "eval_num_tokens": 1261365.0,
+      "eval_runtime": 86.2299,
+      "eval_samples_per_second": 15.957,
+      "eval_steps_per_second": 1.995,
+      "step": 540
+    },
+    {
+      "entropy": 0.5739392962306737,
+      "epoch": 1.3935242839352429,
+      "grad_norm": 0.7475653886795044,
+      "learning_rate": 0.00020061811464663464,
+      "loss": 0.5189042091369629,
+      "mean_token_accuracy": 0.8492388024926185,
+      "num_tokens": 1306879.0,
+      "step": 560
+    },
+    {
+      "epoch": 1.3935242839352429,
+      "eval_entropy": 0.6116398271433142,
+      "eval_loss": 0.551732063293457,
+      "eval_mean_token_accuracy": 0.8450756967067719,
+      "eval_num_tokens": 1306879.0,
+      "eval_runtime": 86.6081,
+      "eval_samples_per_second": 15.888,
+      "eval_steps_per_second": 1.986,
+      "step": 560
+    },
+    {
+      "entropy": 0.5755622573196888,
+      "epoch": 1.4433374844333748,
+      "grad_norm": 0.8218411803245544,
+      "learning_rate": 0.00020036521975103286,
+      "loss": 0.5106248378753662,
+      "mean_token_accuracy": 0.8506785586476326,
+      "num_tokens": 1353534.0,
+      "step": 580
+    },
+    {
+      "epoch": 1.4433374844333748,
+      "eval_entropy": 0.5906928708386976,
+      "eval_loss": 0.551278829574585,
+      "eval_mean_token_accuracy": 0.8462819308042526,
+      "eval_num_tokens": 1353534.0,
+      "eval_runtime": 86.5438,
+      "eval_samples_per_second": 15.899,
+      "eval_steps_per_second": 1.987,
+      "step": 580
+    },
+    {
+      "entropy": 0.5694822132587433,
+      "epoch": 1.4931506849315068,
+      "grad_norm": 0.8880652189254761,
+      "learning_rate": 0.00020008229027548475,
+      "loss": 0.5140334606170655,
+      "mean_token_accuracy": 0.8521522797644139,
+      "num_tokens": 1399537.0,
+      "step": 600
+    },
+    {
+      "epoch": 1.4931506849315068,
+      "eval_entropy": 0.5599641964532608,
+      "eval_loss": 0.5501875877380371,
+      "eval_mean_token_accuracy": 0.8467660788879838,
+      "eval_num_tokens": 1399537.0,
+      "eval_runtime": 86.6458,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 600
+    },
+    {
+      "entropy": 0.5675108034163714,
+      "epoch": 1.5429638854296388,
+      "grad_norm": 0.837087094783783,
+      "learning_rate": 0.0001997694115476612,
+      "loss": 0.5099846363067627,
+      "mean_token_accuracy": 0.8543680295348167,
+      "num_tokens": 1448422.0,
+      "step": 620
+    },
+    {
+      "epoch": 1.5429638854296388,
+      "eval_entropy": 0.5728072581249614,
+      "eval_loss": 0.5445425510406494,
+      "eval_mean_token_accuracy": 0.8474342175001321,
+      "eval_num_tokens": 1448422.0,
+      "eval_runtime": 86.4859,
+      "eval_samples_per_second": 15.91,
+      "eval_steps_per_second": 1.989,
+      "step": 620
+    },
+    {
+      "entropy": 0.5700885068625212,
+      "epoch": 1.592777085927771,
+      "grad_norm": 0.6598765850067139,
+      "learning_rate": 0.000199426677927519,
+      "loss": 0.5122694969177246,
+      "mean_token_accuracy": 0.8519927568733692,
+      "num_tokens": 1495009.0,
+      "step": 640
+    },
+    {
+      "epoch": 1.592777085927771,
+      "eval_entropy": 0.5476993622128353,
+      "eval_loss": 0.5427973866462708,
+      "eval_mean_token_accuracy": 0.8478512147138285,
+      "eval_num_tokens": 1495009.0,
+      "eval_runtime": 86.4172,
+      "eval_samples_per_second": 15.923,
+      "eval_steps_per_second": 1.99,
+      "step": 640
+    },
+    {
+      "entropy": 0.5829229176044464,
+      "epoch": 1.6425902864259028,
+      "grad_norm": 0.6965194940567017,
+      "learning_rate": 0.00019905419277884342,
+      "loss": 0.5253659725189209,
+      "mean_token_accuracy": 0.8493309423327446,
+      "num_tokens": 1536932.0,
+      "step": 660
+    },
+    {
+      "epoch": 1.6425902864259028,
+      "eval_entropy": 0.5666290084983028,
+      "eval_loss": 0.5467478036880493,
+      "eval_mean_token_accuracy": 0.8479407703460649,
+      "eval_num_tokens": 1536932.0,
+      "eval_runtime": 86.4414,
+      "eval_samples_per_second": 15.918,
+      "eval_steps_per_second": 1.99,
+      "step": 660
+    },
+    {
+      "entropy": 0.5498311135917902,
+      "epoch": 1.692403486924035,
+      "grad_norm": 0.636583685874939,
+      "learning_rate": 0.00019865206843807482,
+      "loss": 0.49981012344360354,
+      "mean_token_accuracy": 0.8560848504304885,
+      "num_tokens": 1585718.0,
+      "step": 680
+    },
+    {
+      "epoch": 1.692403486924035,
+      "eval_entropy": 0.539117265406043,
+      "eval_loss": 0.53994220495224,
+      "eval_mean_token_accuracy": 0.8488582601380903,
+      "eval_num_tokens": 1585718.0,
+      "eval_runtime": 86.5296,
+      "eval_samples_per_second": 15.902,
+      "eval_steps_per_second": 1.988,
+      "step": 680
+    },
+    {
+      "entropy": 0.5543891470879316,
+      "epoch": 1.7422166874221667,
+      "grad_norm": 0.6068442463874817,
+      "learning_rate": 0.0001982204261804297,
+      "loss": 0.498047399520874,
+      "mean_token_accuracy": 0.8554679051041603,
+      "num_tokens": 1635718.0,
+      "step": 700
+    },
+    {
+      "epoch": 1.7422166874221667,
+      "eval_entropy": 0.5703774151760478,
+      "eval_loss": 0.5300245881080627,
+      "eval_mean_token_accuracy": 0.850798153946566,
+      "eval_num_tokens": 1635718.0,
+      "eval_runtime": 86.6456,
+      "eval_samples_per_second": 15.881,
+      "eval_steps_per_second": 1.985,
+      "step": 700
+    },
+    {
+      "entropy": 0.546524541825056,
+      "epoch": 1.792029887920299,
+      "grad_norm": 0.7274155020713806,
+      "learning_rate": 0.00019775939618332566,
+      "loss": 0.4988589286804199,
+      "mean_token_accuracy": 0.853422473371029,
+      "num_tokens": 1681291.0,
+      "step": 720
+    },
+    {
+      "epoch": 1.792029887920299,
+      "eval_entropy": 0.5614905688305234,
+      "eval_loss": 0.5350332260131836,
+      "eval_mean_token_accuracy": 0.8492204359797544,
+      "eval_num_tokens": 1681291.0,
+      "eval_runtime": 86.7581,
+      "eval_samples_per_second": 15.86,
+      "eval_steps_per_second": 1.983,
+      "step": 720
+    },
+    {
+      "entropy": 0.5519792139530182,
+      "epoch": 1.841843088418431,
+      "grad_norm": 0.663466215133667,
+      "learning_rate": 0.00019726911748712167,
+      "loss": 0.5099314212799072,
+      "mean_token_accuracy": 0.848412600159645,
+      "num_tokens": 1729102.0,
+      "step": 740
+    },
+    {
+      "epoch": 1.841843088418431,
+      "eval_entropy": 0.5583519090053647,
+      "eval_loss": 0.530483603477478,
+      "eval_mean_token_accuracy": 0.8500003374593202,
+      "eval_num_tokens": 1729102.0,
+      "eval_runtime": 86.3961,
+      "eval_samples_per_second": 15.927,
+      "eval_steps_per_second": 1.991,
+      "step": 740
+    },
+    {
+      "entropy": 0.5454779766499996,
+      "epoch": 1.891656288916563,
+      "grad_norm": 0.890394926071167,
+      "learning_rate": 0.00019674973795318548,
+      "loss": 0.4931994915008545,
+      "mean_token_accuracy": 0.8540832489728928,
+      "num_tokens": 1773578.0,
+      "step": 760
+    },
+    {
+      "epoch": 1.891656288916563,
+      "eval_entropy": 0.572755502406941,
+      "eval_loss": 0.5415747761726379,
+      "eval_mean_token_accuracy": 0.8444425803284312,
+      "eval_num_tokens": 1773578.0,
+      "eval_runtime": 86.4323,
+      "eval_samples_per_second": 15.92,
+      "eval_steps_per_second": 1.99,
+      "step": 760
+    },
+    {
+      "entropy": 0.5392089951783419,
+      "epoch": 1.9414694894146949,
+      "grad_norm": 0.632411777973175,
+      "learning_rate": 0.00019620141421930058,
+      "loss": 0.4957888603210449,
+      "mean_token_accuracy": 0.8549866065382957,
+      "num_tokens": 1821725.0,
+      "step": 780
+    },
+    {
+      "epoch": 1.9414694894146949,
+      "eval_entropy": 0.540764772961306,
+      "eval_loss": 0.5327216386795044,
+      "eval_mean_token_accuracy": 0.850631088364956,
+      "eval_num_tokens": 1821725.0,
+      "eval_runtime": 86.8097,
+      "eval_samples_per_second": 15.851,
+      "eval_steps_per_second": 1.981,
+      "step": 780
+    },
+    {
+      "entropy": 0.5674678739160299,
+      "epoch": 1.9912826899128269,
+      "grad_norm": 0.6958843469619751,
+      "learning_rate": 0.0001956243116524263,
+      "loss": 0.504389762878418,
+      "mean_token_accuracy": 0.8527948908507824,
+      "num_tokens": 1868431.0,
+      "step": 800
+    },
+    {
+      "epoch": 1.9912826899128269,
+      "eval_entropy": 0.530262403190136,
+      "eval_loss": 0.5308871865272522,
+      "eval_mean_token_accuracy": 0.8522498046242913,
+      "eval_num_tokens": 1868431.0,
+      "eval_runtime": 86.7942,
+      "eval_samples_per_second": 15.854,
+      "eval_steps_per_second": 1.982,
+      "step": 800
+    },
+    {
+      "entropy": 0.4742849511213792,
+      "epoch": 2.0398505603985058,
+      "grad_norm": 0.6941492557525635,
+      "learning_rate": 0.00019501860429882556,
+      "loss": 0.418599271774292,
+      "mean_token_accuracy": 0.8748210859604371,
+      "num_tokens": 1915280.0,
+      "step": 820
+    },
+    {
+      "epoch": 2.0398505603985058,
+      "eval_entropy": 0.504602165069691,
+      "eval_loss": 0.542878270149231,
+      "eval_mean_token_accuracy": 0.8507604484641275,
+      "eval_num_tokens": 1915280.0,
+      "eval_runtime": 86.7841,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 820
+    },
+    {
+      "entropy": 0.45857742577791216,
+      "epoch": 2.0896637608966375,
+      "grad_norm": 0.5791997909545898,
+      "learning_rate": 0.00019438447483157478,
+      "loss": 0.399777889251709,
+      "mean_token_accuracy": 0.8754058346152306,
+      "num_tokens": 1965306.0,
+      "step": 840
+    },
+    {
+      "epoch": 2.0896637608966375,
+      "eval_entropy": 0.5028848362176918,
+      "eval_loss": 0.5356478095054626,
+      "eval_mean_token_accuracy": 0.8525635412959165,
+      "eval_num_tokens": 1965306.0,
+      "eval_runtime": 86.6707,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 1.985,
+      "step": 840
+    },
+    {
+      "entropy": 0.4869446292519569,
+      "epoch": 2.1394769613947697,
+      "grad_norm": 0.6483516693115234,
+      "learning_rate": 0.00019372211449547223,
+      "loss": 0.40715818405151366,
+      "mean_token_accuracy": 0.875113020837307,
+      "num_tokens": 2008562.0,
+      "step": 860
+    },
+    {
+      "epoch": 2.1394769613947697,
+      "eval_entropy": 0.4928991326759028,
+      "eval_loss": 0.5419561862945557,
+      "eval_mean_token_accuracy": 0.8516040146350861,
+      "eval_num_tokens": 2008562.0,
+      "eval_runtime": 87.0686,
+      "eval_samples_per_second": 15.804,
+      "eval_steps_per_second": 1.975,
+      "step": 860
+    },
+    {
+      "entropy": 0.45819590501487256,
+      "epoch": 2.1892901618929015,
+      "grad_norm": 0.6661920547485352,
+      "learning_rate": 0.00019303172304936108,
+      "loss": 0.39511430263519287,
+      "mean_token_accuracy": 0.8780680045485496,
+      "num_tokens": 2056474.0,
+      "step": 880
+    },
+    {
+      "epoch": 2.1892901618929015,
+      "eval_entropy": 0.48602560647698334,
+      "eval_loss": 0.5436084866523743,
+      "eval_mean_token_accuracy": 0.8500938470973525,
+      "eval_num_tokens": 2056474.0,
+      "eval_runtime": 86.6809,
+      "eval_samples_per_second": 15.874,
+      "eval_steps_per_second": 1.984,
+      "step": 880
+    },
+    {
+      "entropy": 0.4780638810247183,
+      "epoch": 2.2391033623910337,
+      "grad_norm": 0.6870484352111816,
+      "learning_rate": 0.0001923135087058851,
+      "loss": 0.4061615467071533,
+      "mean_token_accuracy": 0.8766494184732437,
+      "num_tokens": 2103543.0,
+      "step": 900
+    },
+    {
+      "epoch": 2.2391033623910337,
+      "eval_entropy": 0.48236206035281337,
+      "eval_loss": 0.5446090698242188,
+      "eval_mean_token_accuracy": 0.8507725513258646,
+      "eval_num_tokens": 2103543.0,
+      "eval_runtime": 86.7398,
+      "eval_samples_per_second": 15.864,
+      "eval_steps_per_second": 1.983,
+      "step": 900
+    },
+    {
+      "entropy": 0.463029869645834,
+      "epoch": 2.2889165628891655,
+      "grad_norm": 0.6894590854644775,
+      "learning_rate": 0.00019156768806869427,
+      "loss": 0.39602413177490237,
+      "mean_token_accuracy": 0.876420046389103,
+      "num_tokens": 2147861.0,
+      "step": 920
+    },
+    {
+      "epoch": 2.2889165628891655,
+      "eval_entropy": 0.4904779093556626,
+      "eval_loss": 0.5404934287071228,
+      "eval_mean_token_accuracy": 0.852238280828609,
+      "eval_num_tokens": 2147861.0,
+      "eval_runtime": 86.5348,
+      "eval_samples_per_second": 15.901,
+      "eval_steps_per_second": 1.988,
+      "step": 920
+    },
+    {
+      "entropy": 0.4817025110125542,
+      "epoch": 2.3387297633872977,
+      "grad_norm": 0.7756227254867554,
+      "learning_rate": 0.00019079448606712033,
+      "loss": 0.4177968502044678,
+      "mean_token_accuracy": 0.8712256088852882,
+      "num_tokens": 2190561.0,
+      "step": 940
+    },
+    {
+      "epoch": 2.3387297633872977,
+      "eval_entropy": 0.5153802815218305,
+      "eval_loss": 0.5424937605857849,
+      "eval_mean_token_accuracy": 0.8506565759348315,
+      "eval_num_tokens": 2190561.0,
+      "eval_runtime": 86.8973,
+      "eval_samples_per_second": 15.835,
+      "eval_steps_per_second": 1.979,
+      "step": 940
+    },
+    {
+      "entropy": 0.46456389091908934,
+      "epoch": 2.3885429638854294,
+      "grad_norm": 1.2000319957733154,
+      "learning_rate": 0.00018999413588834105,
+      "loss": 0.4084665775299072,
+      "mean_token_accuracy": 0.8750658087432385,
+      "num_tokens": 2239412.0,
+      "step": 960
+    },
+    {
+      "epoch": 2.3885429638854294,
+      "eval_entropy": 0.4849439303195754,
+      "eval_loss": 0.545662522315979,
+      "eval_mean_token_accuracy": 0.8491013112456299,
+      "eval_num_tokens": 2239412.0,
+      "eval_runtime": 86.9049,
+      "eval_samples_per_second": 15.833,
+      "eval_steps_per_second": 1.979,
+      "step": 960
+    },
+    {
+      "entropy": 0.4857471022754908,
+      "epoch": 2.4383561643835616,
+      "grad_norm": 0.9696341753005981,
+      "learning_rate": 0.0001891668789070541,
+      "loss": 0.4149796962738037,
+      "mean_token_accuracy": 0.8704176343977451,
+      "num_tokens": 2286283.0,
+      "step": 980
+    },
+    {
+      "epoch": 2.4383561643835616,
+      "eval_entropy": 0.4872790058684904,
+      "eval_loss": 0.5412707924842834,
+      "eval_mean_token_accuracy": 0.8509329602468846,
+      "eval_num_tokens": 2286283.0,
+      "eval_runtime": 86.7846,
+      "eval_samples_per_second": 15.855,
+      "eval_steps_per_second": 1.982,
+      "step": 980
+    }
+  ],
+  "logging_steps": 20,
+  "max_steps": 4020,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 20,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 9.654290717400883e+16,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}